Add explicit LLM concurrency controls

2026-05-12 21:17:35 +00:00
parent a48f6da1f4
commit 509436cc4a
19 changed files with 875 additions and 99 deletions
--- a/README.md
+++ b/README.md
@@ -121,7 +121,6 @@ Environment:
 - `AUDITA_BASE_URL`
 - `AUDITA_LLM_TIMEOUT_SECONDS`
 - `AUDITA_MAX_RETRIES`
- `AUDITA_LLM_CONCURRENCY`

 CLI:
 - `--llm-api-key`
@@ -129,7 +128,6 @@ CLI:
 - `--base-url`
 - `--llm-timeout-seconds`
 - `--max-retries`
- `--llm-concurrency`

 ### Validation LLM

@@ -151,9 +149,26 @@ CLI:
 - `--validation-llm-concurrency`
 - `--validation-max-prompt-tokens`

-Validation concurrency behavior:
- when validation concurrency is unset, it inherits primary `llm-concurrency`
- when explicitly set, validation concurrency must be `<= llm-concurrency`
+### LLM Concurrency
+
+Environment:
+- `AUDITA_TOTAL_LLM_CONCURRENCY`
+- `AUDITA_PROPOSAL_LLM_CONCURRENCY`
+- `AUDITA_VALIDATION_LLM_CONCURRENCY`
+- `AUDITA_LLM_CONCURRENCY` (legacy alias for `AUDITA_TOTAL_LLM_CONCURRENCY`)
+
+CLI:
+- `--total-llm-concurrency`
+- `--proposal-llm-concurrency`
+- `--validation-llm-concurrency`
+- `--llm-concurrency` (legacy alias for `--total-llm-concurrency`)
+
+Behavior:
+- all proposal and validation LLM calls are bounded by total LLM concurrency
+- proposal LLM calls are additionally bounded by proposal LLM concurrency
+- when validation concurrency is unset, it inherits total LLM concurrency
+- when explicitly set, proposal and validation concurrency must each be `<= total-llm-concurrency`
+- canonical total settings win when both canonical and legacy alias settings are provided at the same precedence layer

 ### Confidence Thresholds

--- a/docs/architecture.md
+++ b/docs/architecture.md
@@ -22,7 +22,7 @@ Implemented today:
 - Deterministic validator-chain execution in the runner with cardinality enforcement.
 - Module-level validator decision/rejection reporting.
 - Internal structured LLM client contract plus an `instructor-go`-backed adapter package.
- Bounded LLM scheduler/semaphore infrastructure with context-aware permit handling.
+- Bounded FIFO LLM scheduler infrastructure with context-aware permit handling.
 - Runtime primary/validation LLM effective-config resolution helpers with validation inheritance.
 - Generic JSON prompt/response diagnostics writer primitives with secret redaction.
 - LLM-backed validator models, prompt builders, batching, and runtime execution.
@@ -162,6 +162,7 @@ Current runtime flow (`internal/cli/run.go`):
    - default run path uses configured default sequence (`glossary,homophones,glossary,spoken_word,grammar`);
    - explicit `--modules` overrides the default sequence;
    - test/injected module factory path remains available for deterministic runtime tests.
+    - each module recomputes chunks from the current working transcript, runs chunk proposal work concurrently, aggregates deterministically, validates, and applies approved proposals once.
 13. Output working transcript to `--output` file or stdout.
 14. Build process report (`phase` currently set to `default_pipeline`).
 15. Optionally write `--report-json`; always write run-dir `report.json`.
@@ -219,6 +220,7 @@ Precedence:
 Implemented config surfaces include:
 - module list
 - primary and validation LLM settings
+- total/proposal/validation LLM concurrency controls
 - section token controls and target sections
 - confidence thresholds
 - normalization controls
@@ -246,10 +248,17 @@ Current runtime boundary:
 - normal `go test ./...` does not require real LLM credentials or Python dependencies.

 `internal/framework/llm` also provides:
- a bounded `Scheduler` for controlled concurrent LLM calls with reliable permit release;
- primary/validation effective-config resolution helpers, including validation inheritance fallback to primary settings;
+- a bounded FIFO `Scheduler` for controlled concurrent LLM calls with reliable permit release on success, error, and cancellation;
+- primary/validation effective-config resolution helpers, including validation inheritance fallback to total LLM concurrency settings;
 - generic interaction diagnostics primitives that write machine-readable JSON artifacts for request metadata, request payload, response payload, and optional error payload with secret redaction.

+LLM concurrency runtime behavior:
+- `total` concurrency bounds all proposal and validation LLM calls.
+- `proposal` concurrency adds a proposal-only sub-cap, composed with total.
+- `validation` concurrency adds a validation-only sub-cap, composed with total.
+- legacy `llm-concurrency` inputs remain compatibility aliases for total concurrency.
+- modules execute serially, chunk proposals run concurrently within each module, and approved proposals are applied once per module in deterministic order.
+
 ## Implemented normalization behavior
 Normalization (`internal/core/normalization`) currently:
 - sorts by segment start time;
--- a/docs/llm-concurrency-audit.md
+++ b/docs/llm-concurrency-audit.md
@@ -20,6 +20,15 @@ Main gap versus the requested target architecture:
 - global concurrency is currently represented by existing `--llm-concurrency`,
 - scheduler implementation is semaphore-based and does not explicitly guarantee FIFO ordering.

+## Implementation Status (2026-05-12 Update)
+
+The targeted concurrency gaps identified in this audit have now been addressed:
+- explicit total/proposal/validation LLM concurrency controls are implemented in config/env/CLI,
+- legacy `llm-concurrency` settings are preserved as compatibility aliases to total concurrency,
+- proposal and validation schedulers are composed with a global total-cap scheduler,
+- scheduler default behavior is FIFO with context-aware queued cancellation and reliable permit release,
+- runner proposal worker fan-out is aligned with effective proposal concurrency.
+
 ## Audit Findings (Questions 1-14)

 1. **Does the current runner execute modules serially?**
--- a/docs/subprocess-operations.md
+++ b/docs/subprocess-operations.md
@@ -16,6 +16,7 @@ audita process <transcript.json> \
 Recommended additions:
 - `--work-dir <dir>` to control diagnostics location.
 - `--work-dir-retention <always|auto|never>` to control retained run directories.
+- `--total-llm-concurrency`, `--proposal-llm-concurrency`, and `--validation-llm-concurrency` when orchestration needs explicit LLM throughput controls.
 - `--modules ...` only when intentionally overriding the default full sequence.

 ## Stdout behavior
--- a/internal/cli/run.go
+++ b/internal/cli/run.go
@@ -196,8 +196,7 @@ var processRunner = func(inv processInvocation, stdout io.Writer) (*normalizatio
 			}
 			globalScheduler := proposalScheduler
 			if globalScheduler == nil {
-				primaryCfg := llm.ResolvePrimaryConfig(inv.Config)
-				s, sErr := llm.NewScheduler(primaryCfg.Concurrency)
+				s, sErr := llm.NewScheduler(inv.Config.TotalLLMConcurrency)
 				if sErr != nil {
 					return fail("runner_setup", sErr, nil)
 				}
@@ -205,12 +204,18 @@ var processRunner = func(inv processInvocation, stdout io.Writer) (*normalizatio
 			}
 			if proposalScheduler == nil {
 				proposalScheduler = globalScheduler
+				if inv.Config.EffectiveProposalLLMConcurrency() < inv.Config.TotalLLMConcurrency {
+					s, sErr := llm.NewScheduler(inv.Config.EffectiveProposalLLMConcurrency())
+					if sErr != nil {
+						return fail("runner_setup", sErr, nil)
+					}
+					proposalScheduler = composeSchedulers(globalScheduler, s)
+				}
 			}
 			if validationScheduler == nil {
 				validationScheduler = globalScheduler
-				if inv.Config.ValidationLLM.Concurrency != nil && inv.Config.PrimaryLLM.Concurrency > inv.Config.EffectiveValidationLLMConfig().Concurrency {
-					validationCfg := llm.ResolveValidationConfig(inv.Config)
-					s, sErr := llm.NewScheduler(validationCfg.Concurrency)
+				if inv.Config.ValidationLLMConcurrency != nil && inv.Config.EffectiveValidationLLMConcurrency() < inv.Config.TotalLLMConcurrency {
+					s, sErr := llm.NewScheduler(inv.Config.EffectiveValidationLLMConcurrency())
 					if sErr != nil {
 						return fail("runner_setup", sErr, nil)
 					}
@@ -392,6 +397,10 @@ func runProcess(args []string, stdout, stderr io.Writer) int {
 			overrides.ValidationBaseURL = pFlags.validationBaseURL
 		case "llm-timeout-seconds":
 			overrides.PrimaryLLMTimeoutSeconds = pFlags.llmTimeoutSeconds
+		case "total-llm-concurrency":
+			overrides.TotalLLMConcurrency = pFlags.totalLLMConcurrency
+		case "proposal-llm-concurrency":
+			overrides.ProposalLLMConcurrency = pFlags.proposalLLMConcurrency
 		case "llm-concurrency":
 			overrides.PrimaryLLMConcurrency = pFlags.llmConcurrency
 		case "validation-llm-timeout-seconds":
@@ -682,6 +691,8 @@ type processFlags struct {
 	baseURL                       *string
 	validationBaseURL             *string
 	llmTimeoutSeconds             *int
+	totalLLMConcurrency           *int
+	proposalLLMConcurrency        *int
 	llmConcurrency                *int
 	validationLLMTimeoutSeconds   *int
 	validationMaxPromptTokens     *int
@@ -717,9 +728,9 @@ func newProcessFlagSet(cfg config.Config, stderr io.Writer) (*flag.FlagSet, proc
 		validationMaxRetriesDefault = *cfg.ValidationLLM.MaxRetries
 	}

-	validationLLMConcurrencyDefault := cfg.PrimaryLLM.Concurrency
-	if cfg.ValidationLLM.Concurrency != nil {
-		validationLLMConcurrencyDefault = *cfg.ValidationLLM.Concurrency
+	validationLLMConcurrencyDefault := cfg.TotalLLMConcurrency
+	if cfg.ValidationLLMConcurrency != nil {
+		validationLLMConcurrencyDefault = *cfg.ValidationLLMConcurrency
 	}

 	targetSectionsDefault := 0
@@ -739,13 +750,15 @@ func newProcessFlagSet(cfg config.Config, stderr io.Writer) (*flag.FlagSet, proc
 		baseURL:                       fs.String("base-url", cfg.PrimaryLLM.BaseURL, "Primary OpenAI-compatible base URL"),
 		validationBaseURL:             fs.String("validation-base-url", cfg.ValidationLLM.BaseURL, "Validation OpenAI-compatible base URL"),
 		llmTimeoutSeconds:             fs.Int("llm-timeout-seconds", cfg.PrimaryLLM.TimeoutSeconds, "Primary LLM timeout in seconds"),
-		llmConcurrency:                fs.Int("llm-concurrency", cfg.PrimaryLLM.Concurrency, "Primary LLM concurrency"),
+		totalLLMConcurrency:           fs.Int("total-llm-concurrency", cfg.TotalLLMConcurrency, "Total concurrent LLM calls across proposal and validation"),
+		proposalLLMConcurrency:        fs.Int("proposal-llm-concurrency", cfg.EffectiveProposalLLMConcurrency(), "Concurrent proposal-generation LLM calls"),
+		llmConcurrency:                fs.Int("llm-concurrency", cfg.TotalLLMConcurrency, "Alias for --total-llm-concurrency"),
 		validationLLMTimeoutSeconds:   fs.Int("validation-llm-timeout-seconds", validationTimeoutSecondsDefault, "Validation LLM timeout in seconds"),
 		validationMaxPromptTokens:     fs.Int("validation-max-prompt-tokens", cfg.ValidationMaxPromptTokens, "Validation max prompt tokens"),
 		targetSections:                fs.Int("target-sections", targetSectionsDefault, "Target number of transcript sections"),
 		maxRetries:                    fs.Int("max-retries", cfg.PrimaryLLM.MaxRetries, "Maximum structured-output retries"),
 		validationMaxRetries:          fs.Int("validation-max-retries", validationMaxRetriesDefault, "Validation structured-output retries"),
-		validationLLMConcurrency:      fs.Int("validation-llm-concurrency", validationLLMConcurrencyDefault, "Validation LLM concurrency"),
+		validationLLMConcurrency:      fs.Int("validation-llm-concurrency", validationLLMConcurrencyDefault, "Concurrent validation LLM calls (inherits total when unset)"),
 		maxSectionTokens:              fs.Int("max-section-tokens", cfg.MaxSectionTokens, "Maximum section tokens"),
 		minSectionTokens:              fs.Int("min-section-tokens", cfg.MinSectionTokens, "Minimum section tokens"),
 		glossaryConfidenceThreshold:   fs.Float64("glossary-confidence-threshold", cfg.Thresholds.Glossary, "Glossary confidence threshold"),
--- a/internal/cli/run_test.go
+++ b/internal/cli/run_test.go
@@ -63,6 +63,8 @@ func TestRunProcessHelpListsExpectedFlags(t *testing.T) {
 		"--base-url",
 		"--validation-base-url",
 		"--llm-timeout-seconds",
+		"--total-llm-concurrency",
+		"--proposal-llm-concurrency",
 		"--llm-concurrency",
 		"--validation-llm-timeout-seconds",
 		"--validation-max-prompt-tokens",
@@ -246,7 +248,7 @@ func TestRunProcessCLIOverridesEnvironment(t *testing.T) {
 	}
 }

-func TestRunProcessRejectsValidationConcurrencyAbovePrimaryConcurrency(t *testing.T) {
+func TestRunProcessRejectsValidationConcurrencyAboveTotalConcurrency(t *testing.T) {
 	var stdout bytes.Buffer
 	var stderr bytes.Buffer

@@ -259,7 +261,7 @@ func TestRunProcessRejectsValidationConcurrencyAbovePrimaryConcurrency(t *testin
 		transcriptPath,
 		"--glossary",
 		fixturePath("tiny_glossary.yaml"),
-		"--llm-concurrency",
+		"--total-llm-concurrency",
 		"1",
 		"--validation-llm-concurrency",
 		"2",
@@ -273,12 +275,12 @@ func TestRunProcessRejectsValidationConcurrencyAbovePrimaryConcurrency(t *testin
 	if !strings.Contains(stderr.String(), "invalid CLI configuration") {
 		t.Fatalf("expected invalid CLI configuration error, got %q", stderr.String())
 	}
-	if !strings.Contains(stderr.String(), "validation llm concurrency must be less than or equal to primary llm concurrency") {
-		t.Fatalf("expected validation/primary concurrency error details, got %q", stderr.String())
+	if !strings.Contains(stderr.String(), "validation llm concurrency must be less than or equal to total llm concurrency") {
+		t.Fatalf("expected validation/total concurrency error details, got %q", stderr.String())
 	}
 }

-func TestRunProcessLLMConcurrencyDrivesEffectiveValidationConcurrencyWhenUnset(t *testing.T) {
+func TestRunProcessTotalLLMConcurrencyDrivesEffectiveValidationConcurrencyWhenUnset(t *testing.T) {
 	processModuleFactory = fakeModuleFactory{modules: map[string]contracts.TranscriptModule{
 		"m": fakeModule{
 			key:    "m",
@@ -288,14 +290,96 @@ func TestRunProcessLLMConcurrencyDrivesEffectiveValidationConcurrencyWhenUnset(t
 					if req.Config == nil {
 						t.Fatal("expected config in validation request")
 					}
-					if req.Config.PrimaryLLM.Concurrency != 4 {
-						t.Fatalf("expected primary llm concurrency 4, got %d", req.Config.PrimaryLLM.Concurrency)
+					if req.Config.TotalLLMConcurrency != 4 {
+						t.Fatalf("expected total llm concurrency 4, got %d", req.Config.TotalLLMConcurrency)
 					}
-					if req.Config.ValidationLLM.Concurrency != nil {
-						t.Fatalf("expected validation concurrency unset, got %#v", req.Config.ValidationLLM.Concurrency)
+					if req.Config.ProposalLLMConcurrency != 4 {
+						t.Fatalf("expected proposal llm concurrency to inherit total 4, got %d", req.Config.ProposalLLMConcurrency)
 					}
-					if req.Config.EffectiveValidationLLMConfig().Concurrency != 4 {
-						t.Fatalf("expected inherited effective validation concurrency 4, got %d", req.Config.EffectiveValidationLLMConfig().Concurrency)
+					if req.Config.ValidationLLMConcurrency != nil {
+						t.Fatalf("expected validation concurrency unset, got %#v", req.Config.ValidationLLMConcurrency)
+					}
+					if req.Config.EffectiveValidationLLMConcurrency() != 4 {
+						t.Fatalf("expected inherited effective validation concurrency 4, got %d", req.Config.EffectiveValidationLLMConcurrency())
+					}
+					return validators.Result{ValidatorName: "capture-config", Decisions: nil}, nil
+				}},
+			},
+			proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
+				return nil, nil
+			},
+		},
+	}}
+	t.Cleanup(func() { processModuleFactory = nil })
+
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+	transcriptPath := writeFile(t, "transcript.json", `[
+  {"id":1,"speaker":"Alice","start":0.0,"end":1.0,"text":"Hello"}
+]`)
+
+	exitCode := Run([]string{
+		"process",
+		transcriptPath,
+		"--glossary",
+		fixturePath("tiny_glossary.yaml"),
+		"--modules",
+		"m",
+		"--total-llm-concurrency",
+		"4",
+	}, &stdout, &stderr)
+	if exitCode != 0 {
+		t.Fatalf("expected success, got %d stderr=%q", exitCode, stderr.String())
+	}
+}
+
+func TestRunProcessRejectsProposalConcurrencyAboveTotalConcurrency(t *testing.T) {
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+
+	transcriptPath := writeFile(t, "transcript.json", `[
+  {"id":1,"speaker":"Alice","start":0.0,"end":1.0,"text":"Hello"}
+]`)
+
+	exitCode := Run([]string{
+		"process",
+		transcriptPath,
+		"--glossary",
+		fixturePath("tiny_glossary.yaml"),
+		"--total-llm-concurrency",
+		"1",
+		"--proposal-llm-concurrency",
+		"2",
+	}, &stdout, &stderr)
+	if exitCode == 0 {
+		t.Fatalf("expected nonzero exit code for invalid llm concurrency combination")
+	}
+	if stdout.Len() != 0 {
+		t.Fatalf("expected empty stdout, got %q", stdout.String())
+	}
+	if !strings.Contains(stderr.String(), "invalid CLI configuration") {
+		t.Fatalf("expected invalid CLI configuration error, got %q", stderr.String())
+	}
+	if !strings.Contains(stderr.String(), "proposal llm concurrency must be less than or equal to total llm concurrency") {
+		t.Fatalf("expected proposal/total concurrency error details, got %q", stderr.String())
+	}
+}
+
+func TestRunProcessLegacyLLMConcurrencyAliasSetsTotalAndProposal(t *testing.T) {
+	processModuleFactory = fakeModuleFactory{modules: map[string]contracts.TranscriptModule{
+		"m": fakeModule{
+			key:    "m",
+			policy: proposals.ReplacementPolicyRequireUnique,
+			validators: []contracts.Validator{
+				fakeValidator{name: "capture-config", validateF: func(req contracts.ValidationRequest) (validators.Result, error) {
+					if req.Config == nil {
+						t.Fatal("expected config in validation request")
+					}
+					if req.Config.TotalLLMConcurrency != 3 {
+						t.Fatalf("expected total llm concurrency 3, got %d", req.Config.TotalLLMConcurrency)
+					}
+					if req.Config.ProposalLLMConcurrency != 3 {
+						t.Fatalf("expected proposal llm concurrency inherited from alias, got %d", req.Config.ProposalLLMConcurrency)
 					}
 					return validators.Result{ValidatorName: "capture-config", Decisions: nil}, nil
 				}},
@@ -321,7 +405,62 @@ func TestRunProcessLLMConcurrencyDrivesEffectiveValidationConcurrencyWhenUnset(t
 		"--modules",
 		"m",
 		"--llm-concurrency",
+		"3",
+	}, &stdout, &stderr)
+	if exitCode != 0 {
+		t.Fatalf("expected success, got %d stderr=%q", exitCode, stderr.String())
+	}
+}
+
+func TestRunProcessLLMConcurrencyFlagsOverrideEnvironment(t *testing.T) {
+	processModuleFactory = fakeModuleFactory{modules: map[string]contracts.TranscriptModule{
+		"m": fakeModule{
+			key:    "m",
+			policy: proposals.ReplacementPolicyRequireUnique,
+			validators: []contracts.Validator{
+				fakeValidator{name: "capture-config", validateF: func(req contracts.ValidationRequest) (validators.Result, error) {
+					if req.Config == nil {
+						t.Fatal("expected config in validation request")
+					}
+					if req.Config.TotalLLMConcurrency != 4 {
+						t.Fatalf("expected cli total llm concurrency 4, got %d", req.Config.TotalLLMConcurrency)
+					}
+					if req.Config.ProposalLLMConcurrency != 3 {
+						t.Fatalf("expected cli proposal llm concurrency 3, got %d", req.Config.ProposalLLMConcurrency)
+					}
+					if req.Config.ValidationLLMConcurrency == nil || *req.Config.ValidationLLMConcurrency != 2 {
+						t.Fatalf("expected env validation llm concurrency 2, got %#v", req.Config.ValidationLLMConcurrency)
+					}
+					return validators.Result{ValidatorName: "capture-config", Decisions: nil}, nil
+				}},
+			},
+			proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
+				return nil, nil
+			},
+		},
+	}}
+	t.Cleanup(func() { processModuleFactory = nil })
+	t.Setenv("AUDITA_TOTAL_LLM_CONCURRENCY", "2")
+	t.Setenv("AUDITA_PROPOSAL_LLM_CONCURRENCY", "2")
+	t.Setenv("AUDITA_VALIDATION_LLM_CONCURRENCY", "2")
+
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+	transcriptPath := writeFile(t, "transcript.json", `[
+  {"id":1,"speaker":"Alice","start":0.0,"end":1.0,"text":"Hello"}
+]`)
+
+	exitCode := Run([]string{
+		"process",
+		transcriptPath,
+		"--glossary",
+		fixturePath("tiny_glossary.yaml"),
+		"--modules",
+		"m",
+		"--total-llm-concurrency",
 		"4",
+		"--proposal-llm-concurrency",
+		"3",
 	}, &stdout, &stderr)
 	if exitCode != 0 {
 		t.Fatalf("expected success, got %d stderr=%q", exitCode, stderr.String())
@@ -373,6 +512,60 @@ func TestComposeSchedulersEnforcesStricterSubcap(t *testing.T) {
 	}
 }

+func TestComposedProposalAndValidationSchedulersShareGlobalTotalCap(t *testing.T) {
+	global, err := llm.NewScheduler(2)
+	if err != nil {
+		t.Fatalf("NewScheduler(global): %v", err)
+	}
+	proposalSubcap, err := llm.NewScheduler(2)
+	if err != nil {
+		t.Fatalf("NewScheduler(proposalSubcap): %v", err)
+	}
+	validationSubcap, err := llm.NewScheduler(2)
+	if err != nil {
+		t.Fatalf("NewScheduler(validationSubcap): %v", err)
+	}
+	proposalScheduler := composeSchedulers(global, proposalSubcap)
+	validationScheduler := composeSchedulers(global, validationSubcap)
+
+	var inFlight int32
+	var maxInFlight int32
+	var wg sync.WaitGroup
+	for i := 0; i < 12; i++ {
+		wg.Add(1)
+		go func(idx int) {
+			defer wg.Done()
+			scheduler := proposalScheduler
+			if idx%2 == 1 {
+				scheduler = validationScheduler
+			}
+			runErr := scheduler.Run(context.Background(), func(context.Context) error {
+				current := atomic.AddInt32(&inFlight, 1)
+				for {
+					prior := atomic.LoadInt32(&maxInFlight)
+					if current <= prior || atomic.CompareAndSwapInt32(&maxInFlight, prior, current) {
+						break
+					}
+				}
+				time.Sleep(20 * time.Millisecond)
+				atomic.AddInt32(&inFlight, -1)
+				return nil
+			})
+			if runErr != nil {
+				t.Errorf("scheduler run error: %v", runErr)
+			}
+		}(i)
+	}
+	wg.Wait()
+
+	if maxInFlight > 2 {
+		t.Fatalf("expected combined proposal+validation concurrency <= global total cap (2), got %d", maxInFlight)
+	}
+	if maxInFlight < 2 {
+		t.Fatalf("expected observed combined concurrency of at least 2, got %d", maxInFlight)
+	}
+}
+
 func TestRunProcessReportJSONSuccessIncludesNormalizationSummary(t *testing.T) {
 	var stdout bytes.Buffer
 	var stderr bytes.Buffer
@@ -2694,6 +2887,12 @@ func TestRunProcessWritesRedactedRunMetadataArtifacts(t *testing.T) {
 		fixturePath("tiny_transcript.json"),
 		"--glossary",
 		fixturePath("tiny_glossary.yaml"),
+		"--total-llm-concurrency",
+		"3",
+		"--proposal-llm-concurrency",
+		"2",
+		"--validation-llm-concurrency",
+		"1",
 		"--output",
 		outputPath,
 		"--report-json",
@@ -2736,6 +2935,24 @@ func TestRunProcessWritesRedactedRunMetadataArtifacts(t *testing.T) {
 		t.Fatalf("effective config artifact leaked API key material")
 	}

+	var effectiveConfig struct {
+		TotalLLMConcurrency      int  `json:"TotalLLMConcurrency"`
+		ProposalLLMConcurrency   int  `json:"ProposalLLMConcurrency"`
+		ValidationLLMConcurrency *int `json:"ValidationLLMConcurrency"`
+	}
+	if err := json.Unmarshal(configBytes, &effectiveConfig); err != nil {
+		t.Fatalf("failed to parse effective config metadata: %v", err)
+	}
+	if effectiveConfig.TotalLLMConcurrency != 3 {
+		t.Fatalf("expected total_llm_concurrency=3 in effective config, got %d", effectiveConfig.TotalLLMConcurrency)
+	}
+	if effectiveConfig.ProposalLLMConcurrency != 2 {
+		t.Fatalf("expected proposal_llm_concurrency=2 in effective config, got %d", effectiveConfig.ProposalLLMConcurrency)
+	}
+	if effectiveConfig.ValidationLLMConcurrency == nil || *effectiveConfig.ValidationLLMConcurrency != 1 {
+		t.Fatalf("expected validation_llm_concurrency=1 in effective config, got %#v", effectiveConfig.ValidationLLMConcurrency)
+	}
+
 	var invocation struct {
 		Operation      string   `json:"operation"`
 		TranscriptPath string   `json:"transcript_path"`
--- a/internal/core/config/config.go
+++ b/internal/core/config/config.go
@@ -36,6 +36,9 @@ type Config struct {
 	Modules                   []string
 	PrimaryLLM                LLMConfig
 	ValidationLLM             ValidationLLMConfig
+	TotalLLMConcurrency       int
+	ProposalLLMConcurrency    int
+	ValidationLLMConcurrency  *int
 	ValidationMaxPromptTokens int
 	MaxSectionTokens          int
 	MinSectionTokens          int
@@ -52,7 +55,9 @@ type LLMConfig struct {
 	BaseURL        string
 	TimeoutSeconds int
 	MaxRetries     int
-	Concurrency    int
+	// Concurrency is retained as a backward-compatible alias for
+	// TotalLLMConcurrency.
+	Concurrency int
 }

 type ValidationLLMConfig struct {
@@ -61,7 +66,9 @@ type ValidationLLMConfig struct {
 	BaseURL        string
 	TimeoutSeconds *int
 	MaxRetries     *int
-	Concurrency    *int
+	// Concurrency is retained as a backward-compatible alias for
+	// ValidationLLMConcurrency.
+	Concurrency *int
 }

 type ConfidenceThresholds struct {
@@ -91,6 +98,9 @@ func Default() Config {
 			Concurrency:    DefaultLLMConcurrency,
 		},
 		ValidationLLM:             ValidationLLMConfig{},
+		TotalLLMConcurrency:       DefaultLLMConcurrency,
+		ProposalLLMConcurrency:    DefaultLLMConcurrency,
+		ValidationLLMConcurrency:  nil,
 		ValidationMaxPromptTokens: DefaultValidationMaxPromptTokens,
 		MaxSectionTokens:          DefaultMaxSectionTokens,
 		MinSectionTokens:          DefaultMinSectionTokens,
@@ -146,9 +156,37 @@ func (c Config) EffectiveValidationLLMConfig() LLMConfig {
 	if c.ValidationLLM.MaxRetries != nil {
 		effective.MaxRetries = *c.ValidationLLM.MaxRetries
 	}
-	if c.ValidationLLM.Concurrency != nil {
-		effective.Concurrency = *c.ValidationLLM.Concurrency
-	}
+	effective.Concurrency = c.EffectiveValidationLLMConcurrency()

 	return effective
 }
+
+func (c Config) EffectiveValidationLLMConcurrency() int {
+	if c.ValidationLLMConcurrency != nil {
+		return *c.ValidationLLMConcurrency
+	}
+	return c.TotalLLMConcurrency
+}
+
+func (c Config) EffectiveProposalLLMConcurrency() int {
+	if c.ProposalLLMConcurrency > 0 {
+		return c.ProposalLLMConcurrency
+	}
+	return c.TotalLLMConcurrency
+}
+
+func (c *Config) syncLegacyConcurrencyAliases() {
+	if c == nil {
+		return
+	}
+	c.PrimaryLLM.Concurrency = c.TotalLLMConcurrency
+	c.ValidationLLM.Concurrency = intPtr(c.ValidationLLMConcurrency)
+}
+
+func intPtr(v *int) *int {
+	if v == nil {
+		return nil
+	}
+	x := *v
+	return &x
+}
--- a/internal/core/config/config_test.go
+++ b/internal/core/config/config_test.go
@@ -24,8 +24,17 @@ func TestDefaultConfigValues(t *testing.T) {
 	if cfg.PrimaryLLM.MaxRetries != DefaultMaxRetries {
 		t.Fatalf("unexpected default max retries: %d", cfg.PrimaryLLM.MaxRetries)
 	}
-	if cfg.PrimaryLLM.Concurrency != DefaultLLMConcurrency {
-		t.Fatalf("unexpected default llm concurrency: %d", cfg.PrimaryLLM.Concurrency)
+	if cfg.TotalLLMConcurrency != DefaultLLMConcurrency {
+		t.Fatalf("unexpected default total llm concurrency: %d", cfg.TotalLLMConcurrency)
+	}
+	if cfg.ProposalLLMConcurrency != DefaultLLMConcurrency {
+		t.Fatalf("unexpected default proposal llm concurrency: %d", cfg.ProposalLLMConcurrency)
+	}
+	if cfg.ValidationLLMConcurrency != nil {
+		t.Fatalf("expected validation llm concurrency to be unset by default")
+	}
+	if cfg.PrimaryLLM.Concurrency != cfg.TotalLLMConcurrency {
+		t.Fatalf("expected primary llm concurrency alias to mirror total, got primary=%d total=%d", cfg.PrimaryLLM.Concurrency, cfg.TotalLLMConcurrency)
 	}
 	if cfg.ValidationLLM.TimeoutSeconds != nil {
 		t.Fatalf("expected validation timeout to be unset by default")
@@ -34,7 +43,7 @@ func TestDefaultConfigValues(t *testing.T) {
 		t.Fatalf("expected validation max retries to be unset by default")
 	}
 	if cfg.ValidationLLM.Concurrency != nil {
-		t.Fatalf("expected validation llm concurrency to be unset by default")
+		t.Fatalf("expected legacy validation llm concurrency alias to be unset by default")
 	}
 	if cfg.TargetSections != nil {
 		t.Fatalf("expected target sections to be unset by default")
@@ -56,7 +65,8 @@ func TestLoadFromEnvOverridesAndFallback(t *testing.T) {
 		"AUDITA_BASE_URL":                         "https://api.openai.com/v1",
 		"AUDITA_LLM_TIMEOUT_SECONDS":              "120",
 		"AUDITA_MAX_RETRIES":                      "7",
-		"AUDITA_LLM_CONCURRENCY":                  "6",
+		"AUDITA_TOTAL_LLM_CONCURRENCY":            "6",
+		"AUDITA_PROPOSAL_LLM_CONCURRENCY":         "4",
 		"AUDITA_VALIDATION_LLM_CONCURRENCY":       "2",
 		"AUDITA_VALIDATION_MAX_PROMPT_TOKENS":     "4096",
 		"AUDITA_MAX_SECTION_TOKENS":               "9000",
@@ -92,17 +102,63 @@ func TestLoadFromEnvOverridesAndFallback(t *testing.T) {
 	if cfg.TargetSections == nil || *cfg.TargetSections != 5 {
 		t.Fatalf("unexpected target sections: %#v", cfg.TargetSections)
 	}
+	if cfg.TotalLLMConcurrency != 6 {
+		t.Fatalf("unexpected total llm concurrency: %d", cfg.TotalLLMConcurrency)
+	}
+	if cfg.ProposalLLMConcurrency != 4 {
+		t.Fatalf("unexpected proposal llm concurrency: %d", cfg.ProposalLLMConcurrency)
+	}
+	if cfg.ValidationLLMConcurrency == nil || *cfg.ValidationLLMConcurrency != 2 {
+		t.Fatalf("unexpected validation llm concurrency: %#v", cfg.ValidationLLMConcurrency)
+	}
 	if cfg.PrimaryLLM.Concurrency != 6 {
-		t.Fatalf("unexpected primary llm concurrency: %d", cfg.PrimaryLLM.Concurrency)
+		t.Fatalf("expected primary alias concurrency 6, got %d", cfg.PrimaryLLM.Concurrency)
 	}
 	if cfg.ValidationLLM.Concurrency == nil || *cfg.ValidationLLM.Concurrency != 2 {
-		t.Fatalf("unexpected validation llm concurrency: %#v", cfg.ValidationLLM.Concurrency)
+		t.Fatalf("expected validation alias concurrency 2, got %#v", cfg.ValidationLLM.Concurrency)
 	}
 	if cfg.WorkDirRetention != WorkDirRetentionAlways {
 		t.Fatalf("unexpected work dir retention: %q", cfg.WorkDirRetention)
 	}
 }

+func TestLoadFromEnvLegacyLLMConcurrencyAliasForTotalAndProposal(t *testing.T) {
+	env := map[string]string{
+		"AUDITA_LLM_CONCURRENCY": "5",
+	}
+
+	cfg, err := loadFromLookup(mapLookup(env))
+	if err != nil {
+		t.Fatalf("loadFromLookup returned error: %v", err)
+	}
+
+	if cfg.TotalLLMConcurrency != 5 {
+		t.Fatalf("expected total concurrency from legacy alias, got %d", cfg.TotalLLMConcurrency)
+	}
+	if cfg.ProposalLLMConcurrency != 5 {
+		t.Fatalf("expected proposal concurrency to inherit legacy total, got %d", cfg.ProposalLLMConcurrency)
+	}
+}
+
+func TestLoadFromEnvCanonicalTotalWinsLegacyAlias(t *testing.T) {
+	env := map[string]string{
+		"AUDITA_TOTAL_LLM_CONCURRENCY": "4",
+		"AUDITA_LLM_CONCURRENCY":       "9",
+	}
+
+	cfg, err := loadFromLookup(mapLookup(env))
+	if err != nil {
+		t.Fatalf("loadFromLookup returned error: %v", err)
+	}
+
+	if cfg.TotalLLMConcurrency != 4 {
+		t.Fatalf("expected canonical total to win over legacy alias, got %d", cfg.TotalLLMConcurrency)
+	}
+	if cfg.ProposalLLMConcurrency != 4 {
+		t.Fatalf("expected proposal to inherit canonical total when unset, got %d", cfg.ProposalLLMConcurrency)
+	}
+}
+
 func TestLoadFromEnvUsesAuditaLLMAPIKeyOverFallback(t *testing.T) {
 	env := map[string]string{
 		"AUDITA_LLM_API_KEY": "primary-key",
@@ -127,12 +183,14 @@ func TestApplyCLIOverridesPrecedence(t *testing.T) {
 	model := "cli-model"
 	workDir := "/cli/work"
 	modules := "grammar"
-	llmConcurrency := 5
+	totalLLMConcurrency := 5
+	proposalLLMConcurrency := 3
 	overrides := CLIOverrides{
-		PrimaryModel:          &model,
-		WorkDir:               &workDir,
-		ModulesCSV:            &modules,
-		PrimaryLLMConcurrency: &llmConcurrency,
+		PrimaryModel:           &model,
+		WorkDir:                &workDir,
+		ModulesCSV:             &modules,
+		TotalLLMConcurrency:    &totalLLMConcurrency,
+		ProposalLLMConcurrency: &proposalLLMConcurrency,
 	}

 	if err := cfg.ApplyCLIOverrides(overrides); err != nil {
@@ -148,17 +206,54 @@ func TestApplyCLIOverridesPrecedence(t *testing.T) {
 	if !reflect.DeepEqual(cfg.Modules, []string{"grammar"}) {
 		t.Fatalf("unexpected modules: %#v", cfg.Modules)
 	}
-	if cfg.PrimaryLLM.Concurrency != 5 {
-		t.Fatalf("expected CLI llm concurrency override, got %d", cfg.PrimaryLLM.Concurrency)
+	if cfg.TotalLLMConcurrency != 5 {
+		t.Fatalf("expected CLI total concurrency override, got %d", cfg.TotalLLMConcurrency)
+	}
+	if cfg.ProposalLLMConcurrency != 3 {
+		t.Fatalf("expected CLI proposal concurrency override, got %d", cfg.ProposalLLMConcurrency)
+	}
+}
+
+func TestApplyCLIOverridesLegacyLLMConcurrencyAlias(t *testing.T) {
+	cfg := Default()
+	aliasConcurrency := 6
+
+	if err := cfg.ApplyCLIOverrides(CLIOverrides{PrimaryLLMConcurrency: &aliasConcurrency}); err != nil {
+		t.Fatalf("ApplyCLIOverrides failed: %v", err)
+	}
+
+	if cfg.TotalLLMConcurrency != 6 {
+		t.Fatalf("expected legacy --llm-concurrency alias to set total, got %d", cfg.TotalLLMConcurrency)
+	}
+	if cfg.ProposalLLMConcurrency != 6 {
+		t.Fatalf("expected proposal to inherit aliased total when unset, got %d", cfg.ProposalLLMConcurrency)
+	}
+}
+
+func TestApplyCLIOverridesCanonicalTotalWinsLegacyAlias(t *testing.T) {
+	cfg := Default()
+	canonicalTotal := 4
+	legacyAlias := 9
+
+	if err := cfg.ApplyCLIOverrides(CLIOverrides{TotalLLMConcurrency: &canonicalTotal, PrimaryLLMConcurrency: &legacyAlias}); err != nil {
+		t.Fatalf("ApplyCLIOverrides failed: %v", err)
+	}
+
+	if cfg.TotalLLMConcurrency != 4 {
+		t.Fatalf("expected canonical total concurrency to win, got %d", cfg.TotalLLMConcurrency)
+	}
+	if cfg.ProposalLLMConcurrency != 4 {
+		t.Fatalf("expected proposal to inherit canonical total when proposal is unset, got %d", cfg.ProposalLLMConcurrency)
 	}
 }

 func TestValidationFailures(t *testing.T) {
 	cfg := Default()
 	cfg.PrimaryLLM.TimeoutSeconds = -1
-	cfg.PrimaryLLM.Concurrency = 0
+	cfg.TotalLLMConcurrency = 0
+	cfg.ProposalLLMConcurrency = 0
 	validationConcurrency := 5
-	cfg.ValidationLLM.Concurrency = &validationConcurrency
+	cfg.ValidationLLMConcurrency = &validationConcurrency
 	cfg.ValidationMaxPromptTokens = 0
 	cfg.MaxSectionTokens = 100
 	cfg.MinSectionTokens = 200
@@ -173,8 +268,9 @@ func TestValidationFailures(t *testing.T) {
 	message := err.Error()
 	for _, expected := range []string{
 		"primary llm timeout seconds",
-		"primary llm concurrency",
-		"validation llm concurrency must be less than or equal to primary llm concurrency",
+		"total llm concurrency",
+		"proposal llm concurrency",
+		"validation llm concurrency must be less than or equal to total llm concurrency",
 		"validation max prompt tokens",
 		"min section tokens",
 		"grammar confidence threshold",
@@ -193,7 +289,8 @@ func TestEffectiveValidationLLMInheritance(t *testing.T) {
 	cfg.PrimaryLLM.BaseURL = "https://primary.example/v1"
 	cfg.PrimaryLLM.TimeoutSeconds = 111
 	cfg.PrimaryLLM.MaxRetries = 2
-	cfg.PrimaryLLM.Concurrency = 7
+	cfg.TotalLLMConcurrency = 7
+	cfg.syncLegacyConcurrencyAliases()

 	effective := cfg.EffectiveValidationLLMConfig()
 	if effective.APIKey != "primary-key" || effective.Model != "primary-model" || effective.BaseURL != "https://primary.example/v1" || effective.TimeoutSeconds != 111 || effective.MaxRetries != 2 || effective.Concurrency != 7 {
@@ -208,7 +305,8 @@ func TestEffectiveValidationLLMInheritance(t *testing.T) {
 	cfg.ValidationLLM.TimeoutSeconds = &validationTimeout
 	cfg.ValidationLLM.MaxRetries = &validationRetries
 	validationConcurrency := 4
-	cfg.ValidationLLM.Concurrency = &validationConcurrency
+	cfg.ValidationLLMConcurrency = &validationConcurrency
+	cfg.syncLegacyConcurrencyAliases()

 	effective = cfg.EffectiveValidationLLMConfig()
 	if effective.APIKey != "validation-key" || effective.Model != "validation-model" || effective.BaseURL != "https://validation.example/v1" || effective.TimeoutSeconds != 222 || effective.MaxRetries != 9 || effective.Concurrency != 4 {
@@ -216,34 +314,50 @@ func TestEffectiveValidationLLMInheritance(t *testing.T) {
 	}
 }

-func TestValidationLLMConcurrencyCannotExceedPrimary(t *testing.T) {
+func TestValidationLLMConcurrencyCannotExceedTotal(t *testing.T) {
 	cfg := Default()
-	cfg.PrimaryLLM.Concurrency = 2
+	cfg.TotalLLMConcurrency = 2
+	cfg.ProposalLLMConcurrency = 2
 	validationConcurrency := 3
-	cfg.ValidationLLM.Concurrency = &validationConcurrency
+	cfg.ValidationLLMConcurrency = &validationConcurrency

 	if err := cfg.Validate(); err == nil {
-		t.Fatal("expected validation error when validation llm concurrency exceeds primary")
+		t.Fatal("expected validation error when validation llm concurrency exceeds total")
 	}

 	validationConcurrency = 2
-	cfg.ValidationLLM.Concurrency = &validationConcurrency
+	cfg.ValidationLLMConcurrency = &validationConcurrency
 	if err := cfg.Validate(); err != nil {
 		t.Fatalf("expected equal concurrency to validate, got %v", err)
 	}
 }

-func TestCLIPrimaryLLMConcurrencyOverrideDrivesEffectiveValidationConcurrencyWhenValidationUnset(t *testing.T) {
+func TestProposalLLMConcurrencyCannotExceedTotal(t *testing.T) {
 	cfg := Default()
-	llmConcurrency := 6
-	if err := cfg.ApplyCLIOverrides(CLIOverrides{PrimaryLLMConcurrency: &llmConcurrency}); err != nil {
+	cfg.TotalLLMConcurrency = 2
+	cfg.ProposalLLMConcurrency = 3
+
+	if err := cfg.Validate(); err == nil {
+		t.Fatal("expected validation error when proposal llm concurrency exceeds total")
+	}
+
+	cfg.ProposalLLMConcurrency = 2
+	if err := cfg.Validate(); err != nil {
+		t.Fatalf("expected equal concurrency to validate, got %v", err)
+	}
+}
+
+func TestCLITotalLLMConcurrencyOverrideDrivesEffectiveValidationConcurrencyWhenValidationUnset(t *testing.T) {
+	cfg := Default()
+	totalLLMConcurrency := 6
+	if err := cfg.ApplyCLIOverrides(CLIOverrides{TotalLLMConcurrency: &totalLLMConcurrency}); err != nil {
 		t.Fatalf("ApplyCLIOverrides failed: %v", err)
 	}
-	if cfg.ValidationLLM.Concurrency != nil {
-		t.Fatalf("expected validation concurrency to remain unset, got %#v", cfg.ValidationLLM.Concurrency)
+	if cfg.ValidationLLMConcurrency != nil {
+		t.Fatalf("expected validation concurrency to remain unset, got %#v", cfg.ValidationLLMConcurrency)
 	}
-	if cfg.EffectiveValidationLLMConfig().Concurrency != 6 {
-		t.Fatalf("expected inherited validation concurrency 6, got %d", cfg.EffectiveValidationLLMConfig().Concurrency)
+	if cfg.EffectiveValidationLLMConcurrency() != 6 {
+		t.Fatalf("expected inherited validation concurrency 6, got %d", cfg.EffectiveValidationLLMConcurrency())
 	}
 }

--- a/internal/core/config/env.go
+++ b/internal/core/config/env.go
@@ -68,12 +68,37 @@ func loadFromLookup(lookup func(string) (string, bool)) (Config, error) {
 		}
 		cfg.PrimaryLLM.MaxRetries = value
 	}
+	totalConcurrencySet := false
+	if raw, ok := lookup("AUDITA_TOTAL_LLM_CONCURRENCY"); ok {
+		value, err := parseInt(raw)
+		if err != nil {
+			return Config{}, fmt.Errorf("AUDITA_TOTAL_LLM_CONCURRENCY: %w", err)
+		}
+		cfg.TotalLLMConcurrency = value
+		totalConcurrencySet = true
+	}
 	if raw, ok := lookup("AUDITA_LLM_CONCURRENCY"); ok {
 		value, err := parseInt(raw)
 		if err != nil {
 			return Config{}, fmt.Errorf("AUDITA_LLM_CONCURRENCY: %w", err)
 		}
-		cfg.PrimaryLLM.Concurrency = value
+		if !totalConcurrencySet {
+			cfg.TotalLLMConcurrency = value
+			totalConcurrencySet = true
+		}
+	}
+
+	proposalConcurrencySet := false
+	if raw, ok := lookup("AUDITA_PROPOSAL_LLM_CONCURRENCY"); ok {
+		value, err := parseInt(raw)
+		if err != nil {
+			return Config{}, fmt.Errorf("AUDITA_PROPOSAL_LLM_CONCURRENCY: %w", err)
+		}
+		cfg.ProposalLLMConcurrency = value
+		proposalConcurrencySet = true
+	}
+	if totalConcurrencySet && !proposalConcurrencySet {
+		cfg.ProposalLLMConcurrency = cfg.TotalLLMConcurrency
 	}

 	if raw, ok := lookup("AUDITA_VALIDATION_MAX_RETRIES"); ok {
@@ -88,7 +113,7 @@ func loadFromLookup(lookup func(string) (string, bool)) (Config, error) {
 		if err != nil {
 			return Config{}, fmt.Errorf("AUDITA_VALIDATION_LLM_CONCURRENCY: %w", err)
 		}
-		cfg.ValidationLLM.Concurrency = &value
+		cfg.ValidationLLMConcurrency = &value
 	}

 	if raw, ok := lookup("AUDITA_VALIDATION_MAX_PROMPT_TOKENS"); ok {
@@ -188,6 +213,8 @@ func loadFromLookup(lookup func(string) (string, bool)) (Config, error) {
 		cfg.WorkDirRetention = WorkDirRetention(raw)
 	}

+	cfg.syncLegacyConcurrencyAliases()
+
 	if err := cfg.Validate(); err != nil {
 		return Config{}, err
 	}
--- a/internal/core/config/flags.go
+++ b/internal/core/config/flags.go
@@ -11,6 +11,8 @@ type CLIOverrides struct {
 	PrimaryBaseURL                *string
 	ValidationBaseURL             *string
 	PrimaryLLMTimeoutSeconds      *int
+	TotalLLMConcurrency           *int
+	ProposalLLMConcurrency        *int
 	PrimaryLLMConcurrency         *int
 	ValidationLLMTimeoutSeconds   *int
 	MaxRetries                    *int
@@ -62,8 +64,24 @@ func (c *Config) ApplyCLIOverrides(overrides CLIOverrides) error {
 	if overrides.PrimaryLLMTimeoutSeconds != nil {
 		c.PrimaryLLM.TimeoutSeconds = *overrides.PrimaryLLMTimeoutSeconds
 	}
-	if overrides.PrimaryLLMConcurrency != nil {
-		c.PrimaryLLM.Concurrency = *overrides.PrimaryLLMConcurrency
+	totalConcurrencySet := false
+	if overrides.TotalLLMConcurrency != nil {
+		c.TotalLLMConcurrency = *overrides.TotalLLMConcurrency
+		totalConcurrencySet = true
+	}
+	// Backward-compatible alias: --llm-concurrency maps to total concurrency
+	// only when --total-llm-concurrency is not set in the same CLI invocation.
+	if overrides.PrimaryLLMConcurrency != nil && !totalConcurrencySet {
+		c.TotalLLMConcurrency = *overrides.PrimaryLLMConcurrency
+		totalConcurrencySet = true
+	}
+	proposalConcurrencySet := false
+	if overrides.ProposalLLMConcurrency != nil {
+		c.ProposalLLMConcurrency = *overrides.ProposalLLMConcurrency
+		proposalConcurrencySet = true
+	}
+	if totalConcurrencySet && !proposalConcurrencySet {
+		c.ProposalLLMConcurrency = c.TotalLLMConcurrency
 	}
 	if overrides.ValidationLLMTimeoutSeconds != nil {
 		value := *overrides.ValidationLLMTimeoutSeconds
@@ -78,7 +96,7 @@ func (c *Config) ApplyCLIOverrides(overrides CLIOverrides) error {
 	}
 	if overrides.ValidationLLMConcurrency != nil {
 		value := *overrides.ValidationLLMConcurrency
-		c.ValidationLLM.Concurrency = &value
+		c.ValidationLLMConcurrency = &value
 	}
 	if overrides.ValidationMaxPromptTokens != nil {
 		c.ValidationMaxPromptTokens = *overrides.ValidationMaxPromptTokens
@@ -124,5 +142,7 @@ func (c *Config) ApplyCLIOverrides(overrides CLIOverrides) error {
 		c.WorkDirRetention = WorkDirRetention(*overrides.WorkDirRetention)
 	}

+	c.syncLegacyConcurrencyAliases()
+
 	return c.Validate()
 }
--- a/internal/core/config/validation.go
+++ b/internal/core/config/validation.go
@@ -24,8 +24,14 @@ func (c Config) Validate() error {
 	if c.PrimaryLLM.MaxRetries < 0 {
 		issues = append(issues, "max retries must be zero or greater")
 	}
-	if c.PrimaryLLM.Concurrency <= 0 {
-		issues = append(issues, "primary llm concurrency must be greater than zero")
+	if c.TotalLLMConcurrency <= 0 {
+		issues = append(issues, "total llm concurrency must be greater than zero")
+	}
+	if c.ProposalLLMConcurrency <= 0 {
+		issues = append(issues, "proposal llm concurrency must be greater than zero")
+	}
+	if c.ProposalLLMConcurrency > c.TotalLLMConcurrency {
+		issues = append(issues, "proposal llm concurrency must be less than or equal to total llm concurrency")
 	}

 	if c.ValidationLLM.TimeoutSeconds != nil && *c.ValidationLLM.TimeoutSeconds <= 0 {
@@ -34,11 +40,11 @@ func (c Config) Validate() error {
 	if c.ValidationLLM.MaxRetries != nil && *c.ValidationLLM.MaxRetries < 0 {
 		issues = append(issues, "validation max retries must be zero or greater")
 	}
-	if c.ValidationLLM.Concurrency != nil && *c.ValidationLLM.Concurrency <= 0 {
+	if c.ValidationLLMConcurrency != nil && *c.ValidationLLMConcurrency <= 0 {
 		issues = append(issues, "validation llm concurrency must be greater than zero")
 	}
-	if c.ValidationLLM.Concurrency != nil && *c.ValidationLLM.Concurrency > c.PrimaryLLM.Concurrency {
-		issues = append(issues, "validation llm concurrency must be less than or equal to primary llm concurrency")
+	if c.ValidationLLMConcurrency != nil && *c.ValidationLLMConcurrency > c.TotalLLMConcurrency {
+		issues = append(issues, "validation llm concurrency must be less than or equal to total llm concurrency")
 	}

 	if c.ValidationMaxPromptTokens <= 0 {
--- a/internal/framework/llm/effective_config.go
+++ b/internal/framework/llm/effective_config.go
@@ -21,13 +21,13 @@ type EffectiveConfig struct {

 // ResolvePrimaryConfig resolves the primary LLM settings into runtime form.
 func ResolvePrimaryConfig(cfg config.Config) EffectiveConfig {
-	return resolveFromLLMConfig(cfg.PrimaryLLM)
+	return resolveFromLLMConfig(cfg.PrimaryLLM, cfg.TotalLLMConcurrency)
 }

 // ResolveValidationConfig resolves validation LLM settings with inheritance
 // from primary values when validation overrides are unset.
 func ResolveValidationConfig(cfg config.Config) EffectiveConfig {
-	return resolveFromLLMConfig(cfg.EffectiveValidationLLMConfig())
+	return resolveFromLLMConfig(cfg.EffectiveValidationLLMConfig(), cfg.EffectiveValidationLLMConcurrency())
 }

 // ToInstructorClientConfig converts an effective runtime config into adapter
@@ -44,13 +44,13 @@ func (c EffectiveConfig) ToInstructorClientConfig(mode Mode, httpClient *http.Cl
 	}
 }

-func resolveFromLLMConfig(raw config.LLMConfig) EffectiveConfig {
+func resolveFromLLMConfig(raw config.LLMConfig, concurrency int) EffectiveConfig {
 	return EffectiveConfig{
 		APIKey:         strings.TrimSpace(raw.APIKey),
 		Model:          strings.TrimSpace(raw.Model),
 		BaseURL:        strings.TrimSpace(raw.BaseURL),
 		RequestTimeout: time.Duration(raw.TimeoutSeconds) * time.Second,
 		MaxRetries:     raw.MaxRetries,
-		Concurrency:    raw.Concurrency,
+		Concurrency:    concurrency,
 	}
 }
--- a/internal/framework/llm/effective_config_test.go
+++ b/internal/framework/llm/effective_config_test.go
@@ -14,7 +14,8 @@ func TestResolvePrimaryConfig(t *testing.T) {
 	cfg.PrimaryLLM.BaseURL = " https://example.test/v1 "
 	cfg.PrimaryLLM.TimeoutSeconds = 42
 	cfg.PrimaryLLM.MaxRetries = 7
-	cfg.PrimaryLLM.Concurrency = 3
+	cfg.TotalLLMConcurrency = 3
+	cfg.ProposalLLMConcurrency = 3

 	effective := ResolvePrimaryConfig(cfg)
 	if effective.APIKey != "primary-key" {
@@ -44,7 +45,8 @@ func TestResolveValidationConfigInheritsPrimaryWhenUnset(t *testing.T) {
 	cfg.PrimaryLLM.BaseURL = "https://primary.example/v1"
 	cfg.PrimaryLLM.TimeoutSeconds = 90
 	cfg.PrimaryLLM.MaxRetries = 4
-	cfg.PrimaryLLM.Concurrency = 2
+	cfg.TotalLLMConcurrency = 2
+	cfg.ProposalLLMConcurrency = 2

 	effective := ResolveValidationConfig(cfg)
 	if effective.APIKey != "primary-key" ||
@@ -64,7 +66,8 @@ func TestResolveValidationConfigOverridesPrimaryWhenSet(t *testing.T) {
 	cfg.PrimaryLLM.BaseURL = "https://primary.example/v1"
 	cfg.PrimaryLLM.TimeoutSeconds = 90
 	cfg.PrimaryLLM.MaxRetries = 4
-	cfg.PrimaryLLM.Concurrency = 2
+	cfg.TotalLLMConcurrency = 2
+	cfg.ProposalLLMConcurrency = 2

 	timeout := 12
 	retries := 9
@@ -74,7 +77,7 @@ func TestResolveValidationConfigOverridesPrimaryWhenSet(t *testing.T) {
 	cfg.ValidationLLM.BaseURL = "https://validation.example/v1"
 	cfg.ValidationLLM.TimeoutSeconds = &timeout
 	cfg.ValidationLLM.MaxRetries = &retries
-	cfg.ValidationLLM.Concurrency = &concurrency
+	cfg.ValidationLLMConcurrency = &concurrency

 	effective := ResolveValidationConfig(cfg)
 	if effective.APIKey != "validation-key" ||
--- a/internal/framework/llm/scheduler.go
+++ b/internal/framework/llm/scheduler.go
@@ -8,7 +8,16 @@ import (

 // Scheduler bounds concurrent backend LLM calls.
 type Scheduler struct {
-	permits chan struct{}
+	maxConcurrency int
+	mu             sync.Mutex
+	inFlight       int
+	queue          []*waiter
+}
+
+type waiter struct {
+	ready   chan struct{}
+	queued  bool
+	granted bool
 }

 // NewScheduler creates a scheduler with a fixed concurrency limit.
@@ -17,26 +26,86 @@ func NewScheduler(maxConcurrency int) (*Scheduler, error) {
 		return nil, fmt.Errorf("max concurrency must be greater than zero")
 	}
 	return &Scheduler{
-		permits: make(chan struct{}, maxConcurrency),
+		maxConcurrency: maxConcurrency,
 	}, nil
 }

 // Acquire blocks until a permit is available or the context is canceled.
 // The returned release function is safe to call multiple times.
 func (s *Scheduler) Acquire(ctx context.Context) (func(), error) {
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
+
+	s.mu.Lock()
+	if s.inFlight < s.maxConcurrency && len(s.queue) == 0 {
+		s.inFlight++
+		s.mu.Unlock()
+		return s.releaseFunc(), nil
+	}
+	w := &waiter{
+		ready:  make(chan struct{}),
+		queued: true,
+	}
+	s.queue = append(s.queue, w)
+	s.mu.Unlock()
+
 	select {
-	case s.permits <- struct{}{}:
-		var once sync.Once
-		return func() {
-			once.Do(func() {
-				<-s.permits
-			})
-		}, nil
+	case <-w.ready:
+		return s.releaseFunc(), nil
 	case <-ctx.Done():
+		s.mu.Lock()
+		if w.queued {
+			s.removeQueuedWaiterLocked(w)
+			s.mu.Unlock()
+			return nil, ctx.Err()
+		}
+		if w.granted {
+			// Permit was granted concurrently with cancellation; release it to
+			// avoid leaks and unblock queued callers.
+			s.inFlight--
+			s.grantQueuedLocked()
+		}
+		s.mu.Unlock()
 		return nil, ctx.Err()
 	}
 }

+func (s *Scheduler) releaseFunc() func() {
+	var once sync.Once
+	return func() {
+		once.Do(func() {
+			s.mu.Lock()
+			if s.inFlight > 0 {
+				s.inFlight--
+				s.grantQueuedLocked()
+			}
+			s.mu.Unlock()
+		})
+	}
+}
+
+func (s *Scheduler) grantQueuedLocked() {
+	for s.inFlight < s.maxConcurrency && len(s.queue) > 0 {
+		w := s.queue[0]
+		s.queue = s.queue[1:]
+		w.queued = false
+		w.granted = true
+		s.inFlight++
+		close(w.ready)
+	}
+}
+
+func (s *Scheduler) removeQueuedWaiterLocked(target *waiter) {
+	for i, w := range s.queue {
+		if w == target {
+			w.queued = false
+			s.queue = append(s.queue[:i], s.queue[i+1:]...)
+			return
+		}
+	}
+}
+
 // Run acquires a permit, executes fn, and releases the permit.
 func (s *Scheduler) Run(ctx context.Context, fn func(context.Context) error) error {
 	release, err := s.Acquire(ctx)
--- a/internal/framework/llm/scheduler_test.go
+++ b/internal/framework/llm/scheduler_test.go
@@ -3,12 +3,64 @@ package llm
 import (
 	"context"
 	"errors"
+	"reflect"
 	"sync"
 	"sync/atomic"
 	"testing"
 	"time"
 )

+func TestSchedulerFIFOOrdering(t *testing.T) {
+	s, err := NewScheduler(1)
+	if err != nil {
+		t.Fatalf("NewScheduler: %v", err)
+	}
+
+	firstRelease, err := s.Acquire(context.Background())
+	if err != nil {
+		t.Fatalf("Acquire(first): %v", err)
+	}
+
+	gotOrder := make(chan int, 3)
+	waitChans := []chan struct{}{
+		make(chan struct{}),
+		make(chan struct{}),
+		make(chan struct{}),
+	}
+	var wg sync.WaitGroup
+	for i := 0; i < 3; i++ {
+		id := i
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			runErr := s.Run(context.Background(), func(context.Context) error {
+				gotOrder <- id
+				<-waitChans[id]
+				return nil
+			})
+			if runErr != nil {
+				t.Errorf("Run[%d] error: %v", id, runErr)
+			}
+		}()
+		time.Sleep(10 * time.Millisecond)
+	}
+
+	firstRelease()
+
+	order := make([]int, 0, 3)
+	for i := 0; i < 3; i++ {
+		id := <-gotOrder
+		order = append(order, id)
+		close(waitChans[id])
+	}
+
+	wg.Wait()
+
+	if !reflect.DeepEqual(order, []int{0, 1, 2}) {
+		t.Fatalf("expected FIFO order [0 1 2], got %v", order)
+	}
+}
+
 func TestSchedulerEnforcesMaxConcurrency(t *testing.T) {
 	s, err := NewScheduler(2)
 	if err != nil {
@@ -79,7 +131,6 @@ func TestSchedulerReleasesPermitOnError(t *testing.T) {
 		t.Fatalf("expected %v, got %v", expectedErr, err)
 	}

-	// Must be able to run again after an error, proving permit release.
 	if err := s.Run(context.Background(), func(ctx context.Context) error {
 		_ = ctx
 		return nil
@@ -88,7 +139,7 @@ func TestSchedulerReleasesPermitOnError(t *testing.T) {
 	}
 }

-func TestSchedulerRespectsContextCancellation(t *testing.T) {
+func TestSchedulerContextCancellationWhileQueued(t *testing.T) {
 	s, err := NewScheduler(1)
 	if err != nil {
 		t.Fatalf("NewScheduler: %v", err)
@@ -111,3 +162,27 @@ func TestSchedulerRespectsContextCancellation(t *testing.T) {
 		t.Fatalf("expected deadline exceeded, got %v", err)
 	}
 }
+
+func TestSchedulerNoPermitLeakAfterQueuedCancellation(t *testing.T) {
+	s, err := NewScheduler(1)
+	if err != nil {
+		t.Fatalf("NewScheduler: %v", err)
+	}
+
+	firstRelease, err := s.Acquire(context.Background())
+	if err != nil {
+		t.Fatalf("Acquire(first): %v", err)
+	}
+
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel()
+	if _, err := s.Acquire(ctx); !errors.Is(err, context.Canceled) {
+		t.Fatalf("expected context canceled, got %v", err)
+	}
+
+	firstRelease()
+
+	if err := s.Run(context.Background(), func(context.Context) error { return nil }); err != nil {
+		t.Fatalf("expected scheduler to accept new work after cancellation, got %v", err)
+	}
+}
--- a/internal/framework/proposal_generation/generate_test.go
+++ b/internal/framework/proposal_generation/generate_test.go
@@ -7,10 +7,14 @@ import (
 	"path/filepath"
 	"reflect"
 	"strings"
+	"sync"
+	"sync/atomic"
 	"testing"
+	"time"

 	"gitea.maximumdirect.net/eric/audita/internal/core/config"
 	"gitea.maximumdirect.net/eric/audita/internal/framework/contracts"
+	"gitea.maximumdirect.net/eric/audita/internal/framework/llm"
 	"gitea.maximumdirect.net/eric/audita/internal/framework/proposals"
 )

@@ -47,6 +51,40 @@ func (s *countingScheduler) Run(ctx context.Context, fn func(context.Context) er
 	return fn(ctx)
 }

+type sleepingStructuredClient struct {
+	inFlight    int32
+	maxInFlight int32
+}
+
+func (c *sleepingStructuredClient) CompleteStructured(ctx context.Context, req contracts.StructuredCompletionRequest, out any) (contracts.StructuredCompletionResponse, error) {
+	_ = req
+	current := atomic.AddInt32(&c.inFlight, 1)
+	for {
+		prior := atomic.LoadInt32(&c.maxInFlight)
+		if current <= prior || atomic.CompareAndSwapInt32(&c.maxInFlight, prior, current) {
+			break
+		}
+	}
+	select {
+	case <-time.After(20 * time.Millisecond):
+	case <-ctx.Done():
+		atomic.AddInt32(&c.inFlight, -1)
+		return contracts.StructuredCompletionResponse{}, ctx.Err()
+	}
+	atomic.AddInt32(&c.inFlight, -1)
+
+	target, ok := out.(*StructuredCorrectionSet)
+	if !ok {
+		return contracts.StructuredCompletionResponse{}, errors.New("unexpected output type")
+	}
+	*target = StructuredCorrectionSet{
+		Corrections: []StructuredCorrectionProposal{
+			{TargetSegmentID: 1, OriginalText: "x", CorrectedText: "y", Confidence: 0.9},
+		},
+	}
+	return contracts.StructuredCompletionResponse{}, nil
+}
+
 func defaultRequest(t *testing.T) Request {
 	t.Helper()
 	cfg := config.Default()
@@ -265,3 +303,35 @@ func TestGenerateCandidatesClientError(t *testing.T) {
 		t.Fatalf("expected error diagnostics artifact under %s", req.DiagnosticsDir)
 	}
 }
+
+func TestGenerateCandidatesRespectsSchedulerConcurrency(t *testing.T) {
+	scheduler, err := llm.NewScheduler(2)
+	if err != nil {
+		t.Fatalf("NewScheduler: %v", err)
+	}
+	client := &sleepingStructuredClient{}
+	baseReq := defaultRequest(t)
+	baseReq.LLMClient = client
+	baseReq.Scheduler = scheduler
+
+	var wg sync.WaitGroup
+	for i := 0; i < 8; i++ {
+		wg.Add(1)
+		go func(idx int) {
+			defer wg.Done()
+			req := baseReq
+			req.StartIndex = idx
+			if _, runErr := GenerateCandidates(context.Background(), req); runErr != nil {
+				t.Errorf("GenerateCandidates[%d] error: %v", idx, runErr)
+			}
+		}(i)
+	}
+	wg.Wait()
+
+	if got := atomic.LoadInt32(&client.maxInFlight); got > 2 {
+		t.Fatalf("expected scheduler cap <= 2, got %d", got)
+	}
+	if got := atomic.LoadInt32(&client.maxInFlight); got < 2 {
+		t.Fatalf("expected observed concurrency of at least 2, got %d", got)
+	}
+}
--- a/internal/framework/runner/runner.go
+++ b/internal/framework/runner/runner.go
@@ -296,8 +296,8 @@ func collectSectionProposals(ctx context.Context, input collectSectionProposalsI
 	}

 	maxWorkers := 1
-	if input.Config != nil && input.Config.PrimaryLLM.Concurrency > 1 {
-		maxWorkers = input.Config.PrimaryLLM.Concurrency
+	if input.Config != nil && input.Config.EffectiveProposalLLMConcurrency() > 1 {
+		maxWorkers = input.Config.EffectiveProposalLLMConcurrency()
 	}
 	if maxWorkers > len(input.Sections) {
 		maxWorkers = len(input.Sections)
--- a/internal/framework/runner/runner_test.go
+++ b/internal/framework/runner/runner_test.go
@@ -163,11 +163,12 @@ func TestRunnerProposalsExecutePerChunkSection(t *testing.T) {
 	}
 }

-func TestRunnerProposalSectionConcurrencyBoundedByPrimaryLLMConcurrency(t *testing.T) {
+func TestRunnerProposalSectionConcurrencyBoundedByProposalLLMConcurrency(t *testing.T) {
 	cfg := config.Default()
 	cfg.MaxSectionTokens = 3
 	cfg.MinSectionTokens = 0
-	cfg.PrimaryLLM.Concurrency = 2
+	cfg.TotalLLMConcurrency = 2
+	cfg.ProposalLLMConcurrency = 2

 	transcript := &schema.Transcript{Segments: []schema.Segment{
 		{ID: 1, Text: "one two"},
@@ -215,7 +216,8 @@ func TestRunnerProposalIndexOrderingIsDeterministicAcrossParallelSections(t *tes
 	cfg := config.Default()
 	cfg.MaxSectionTokens = 3
 	cfg.MinSectionTokens = 0
-	cfg.PrimaryLLM.Concurrency = 3
+	cfg.TotalLLMConcurrency = 3
+	cfg.ProposalLLMConcurrency = 3

 	transcript := &schema.Transcript{Segments: []schema.Segment{
 		{ID: 1, Text: "alpha one"},
@@ -259,6 +261,12 @@ func TestRunnerProposalIndexOrderingIsDeterministicAcrossParallelSections(t *tes
 			t.Fatalf("expected deterministic applied-change order by section/segment, got %+v", out.ModuleResults[0].AppliedChanges)
 		}
 	}
+	wantFinal := []string{"ALPHA ONE", "BRAVO TWO", "CHARLIE THREE"}
+	for i, seg := range out.FinalTranscript.Segments {
+		if seg.Text != wantFinal[i] {
+			t.Fatalf("expected deterministic final transcript regardless of section completion order; got %+v", out.FinalTranscript.Segments)
+		}
+	}
 }

 func TestRunnerSkippedRecorded(t *testing.T) {
--- a/internal/framework/validators/llm_validators_test.go
+++ b/internal/framework/validators/llm_validators_test.go
@@ -4,7 +4,10 @@ import (
 	"context"
 	"errors"
 	"strings"
+	"sync"
+	"sync/atomic"
 	"testing"
+	"time"

 	"gitea.maximumdirect.net/eric/audita/internal/core/chunking"
 	"gitea.maximumdirect.net/eric/audita/internal/core/config"
@@ -18,6 +21,54 @@ type fakeStructuredLLMClient struct {
 	calls     []StructuredCompletionRequest
 }

+type sleepingValidationClient struct {
+	inFlight    int32
+	maxInFlight int32
+}
+
+type boundedScheduler struct {
+	permits chan struct{}
+}
+
+func newBoundedScheduler(max int) *boundedScheduler {
+	return &boundedScheduler{permits: make(chan struct{}, max)}
+}
+
+func (s *boundedScheduler) Run(ctx context.Context, fn func(context.Context) error) error {
+	select {
+	case s.permits <- struct{}{}:
+	case <-ctx.Done():
+		return ctx.Err()
+	}
+	defer func() { <-s.permits }()
+	return fn(ctx)
+}
+
+func (c *sleepingValidationClient) CompleteStructured(ctx context.Context, req StructuredCompletionRequest, out any) (StructuredCompletionResponse, error) {
+	_ = req
+	current := atomic.AddInt32(&c.inFlight, 1)
+	for {
+		prior := atomic.LoadInt32(&c.maxInFlight)
+		if current <= prior || atomic.CompareAndSwapInt32(&c.maxInFlight, prior, current) {
+			break
+		}
+	}
+	select {
+	case <-time.After(20 * time.Millisecond):
+	case <-ctx.Done():
+		atomic.AddInt32(&c.inFlight, -1)
+		return StructuredCompletionResponse{}, ctx.Err()
+	}
+	atomic.AddInt32(&c.inFlight, -1)
+	target := out.(*LLMValidationResponse)
+	*target = LLMValidationResponse{
+		Validations: []LLMValidationDecision{
+			{CorrectionIndex: 0, Approved: true, Confidence: 0.9, Reason: "ok"},
+		},
+	}
+	return StructuredCompletionResponse{}, nil
+}
+
 func (f *fakeStructuredLLMClient) CompleteStructured(ctx context.Context, req StructuredCompletionRequest, out any) (StructuredCompletionResponse, error) {
 	_ = ctx
 	f.calls = append(f.calls, req)
@@ -191,3 +242,34 @@ func TestLLMBackedValidatorUnknownProposalIndexFails(t *testing.T) {
 		t.Fatalf("expected unknown index error, got %v", err)
 	}
 }
+
+func TestLLMBackedValidatorRespectsSchedulerConcurrency(t *testing.T) {
+	scheduler := newBoundedScheduler(2)
+	client := &sleepingValidationClient{}
+	v, err := NewLLMBackedValidator("spoken_form_plausibility_review", LLMValidatorTypeSpokenFormPlausibility, "test-model")
+	if err != nil {
+		t.Fatalf("new validator error: %v", err)
+	}
+
+	var wg sync.WaitGroup
+	for i := 0; i < 8; i++ {
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			req := makeReq([]proposals.EnrichedCorrectionProposal{mk(0, "gestures", "Jesters")})
+			req.LLMClient = client
+			req.Scheduler = scheduler
+			if _, runErr := v.Validate(context.Background(), req); runErr != nil {
+				t.Errorf("Validate error: %v", runErr)
+			}
+		}()
+	}
+	wg.Wait()
+
+	if got := atomic.LoadInt32(&client.maxInFlight); got > 2 {
+		t.Fatalf("expected scheduler cap <= 2, got %d", got)
+	}
+	if got := atomic.LoadInt32(&client.maxInFlight); got < 2 {
+		t.Fatalf("expected observed concurrency of at least 2, got %d", got)
+	}
+}