Add persistent job claim and file artifact reads

api/platform_client.go

@@ -8,6 +8,7 @@ import (
 	"crypto/sha256"
 	"encoding/hex"
 	"encoding/json"
+	"errors"
 	"fmt"
 	"io"
 	"log"
@@ -152,7 +153,25 @@ func (c PlatformClient) ReportLifecycle(ctx context.Context, request protocol.Ru
 }
 
 func (c PlatformClient) ClaimJob(ctx context.Context, request protocol.RunJobClaimRequest) (protocol.RunJobClaimResponse, error) {
+	response, err := postPlatformJSON[protocol.RunJobClaimRequest, protocol.RunJobClaimResponse](ctx, c, "/api/v1/run/jobs/claim", request)
+	if request.WaitSeconds > 0 && claimWaitUnsupported(err) {
+		request.WaitSeconds = 0
 		return postPlatformJSON[protocol.RunJobClaimRequest, protocol.RunJobClaimResponse](ctx, c, "/api/v1/run/jobs/claim", request)
+	}
+	return response, err
+}
+
+func claimWaitUnsupported(err error) bool {
+	var requestErr PlatformRequestError
+	if !errors.As(err, &requestErr) || requestErr.Status != http.StatusBadRequest || requestErr.Code != "bad_request" {
+		return false
+	}
+	for _, detail := range requestErr.Details {
+		if strings.Contains(detail, "waitSeconds") {
+			return true
+		}
+	}
+	return false
 }
 
 func (c PlatformClient) AckJob(ctx context.Context, request protocol.RunJobAckRequest) (protocol.RunJobAckResponse, error) {

protocol/job.go

@@ -321,6 +321,7 @@ type RunJobClaimRequest struct {
 	SessionToken  string            `json:"sessionToken"`
 	Capabilities  []string          `json:"capabilities"`
 	Capacity      RunCapacityReport `json:"capacity"`
+	WaitSeconds   int               `json:"waitSeconds,omitempty"`
 }
 
 type RunJobClaimResponse struct {

protocol/job.md

@@ -4,7 +4,7 @@ Jobs execute bounded server management work.
 
 ## Implemented Routes
 
-- `POST /api/v1/run/jobs/claim`: claims one queued job for the registered run endpoint.
+- `POST /api/v1/run/jobs/claim`: claims one queued job for the registered run endpoint, optionally holding the request for a bounded wait window so Platform can wake Run immediately when work arrives.
 - `POST /api/v1/run/jobs/ack`: acknowledges an active leased job before execution.
 - `POST /api/v1/run/jobs/progress`: reports bounded progress for an active leased job.
 - `POST /api/v1/run/jobs/result`: submits a bounded terminal result for an active leased job.
@@ -17,7 +17,7 @@ Jobs execute bounded server management work.
 
 ## Payloads
 
-- `RunJobClaimRequest`: session token, run ID, capacity, and supported capabilities.
+- `RunJobClaimRequest`: session token, run ID, capacity, supported capabilities, and optional `waitSeconds` for long-poll claim waiting.
 - `RunJobClaimResponse`: optional job assignment with identity, capability, server instance, logical target key, scoped input ref, idempotency key, per-job attempt, max attempts, raw one-use lease token, ack deadline, execution lease deadline, and polling hint. Platform persists only the lease hash.
 - `RunJobAckRequest`: job ID, run ID, session token, lease token, attempt, and bounded message.
 - `RunJobProgressRequest`: job ID, run ID, session token, lease token, attempt, percent, sequence, and bounded message.

runtime/execution_test.go

@@ -676,6 +676,36 @@ func TestDeploymentFileExecutorUsesServerRoot(t *testing.T) {
 	}
 }
 
+func TestDeploymentFileExecutorListsNestedPathsFromServerRoot(t *testing.T) {
+	workspaceRoot := t.TempDir()
+	serverRoot := t.TempDir()
+	if err := os.MkdirAll(filepath.Join(serverRoot, "SCUM", "Saved"), 0o700); err != nil {
+		t.Fatalf("mkdir server root fixture: %v", err)
+	}
+	if err := os.WriteFile(filepath.Join(serverRoot, "SCUM", "Saved", "ServerSettings.ini"), []byte("[/Script/SCUM.ServerSettings]\n"), 0o600); err != nil {
+		t.Fatalf("write server root fixture: %v", err)
+	}
+	executor, err := NewFileExecutor(workspaceRoot)
+	if err != nil {
+		t.Fatalf("new file executor: %v", err)
+	}
+	assignment := executionAssignment(protocol.RunCapabilityFilesList)
+	assignment.TargetKey = "server-root"
+	assignment.ExecutionInput.Deployment = &protocol.ServerDeploymentExecution{SchemaVersion: "1", Mode: "guided-install", ServerRoot: serverRoot, Revision: 7}
+	assignment.ExecutionInput.Inputs = map[string]string{"path": "SCUM", "recursive": "true", "query": ""}
+	result := executor.Execute(context.Background(), assignment)
+	if result.State != lifecycleResultStateSucceeded || result.ExecutionResult.Kind != "file.list" {
+		t.Fatalf("expected deployment root listing, got %+v", result)
+	}
+	var envelope fileListEnvelope
+	if err := json.Unmarshal([]byte(result.ExecutionResult.Content), &envelope); err != nil {
+		t.Fatalf("decode listing: %v", err)
+	}
+	if envelope.Path != "SCUM" || len(envelope.Entries) != 2 || envelope.Entries[0].RelativePath != "SCUM/Saved" || envelope.Entries[1].RelativePath != "SCUM/Saved/ServerSettings.ini" {
+		t.Fatalf("expected deployment-root relative paths, got %+v", envelope)
+	}
+}
+
 func TestScopedFileExecutorCancellationLeavesTargetUnchanged(t *testing.T) {
 	root := t.TempDir()
 	executor, err := NewFileExecutor(root)

runtime/file_artifact_transfer.go

@@ -0,0 +1,129 @@
+package runtime
+
+import (
+	"context"
+	"crypto/sha256"
+	"encoding/hex"
+	"errors"
+	"fmt"
+	"io"
+	"os"
+
+	"browser.local/run/protocol"
+)
+
+func (worker *Worker) uploadFileArtifact(ctx context.Context, assignment protocol.RunJobAssignment, artifactID string, targetPath string, sizeBytes int64, checksum string) error {
+	state, err := worker.registeredState()
+	if err != nil {
+		return err
+	}
+	opened, err := worker.client.OpenArtifactTransfer(ctx, protocol.ArtifactTransferOpenRequest{
+		RunEndpointID:  state.RunEndpointID,
+		SessionToken:   state.SessionToken,
+		ArtifactID:     artifactID,
+		Direction:      "upload",
+		OwnerKind:      "job",
+		OwnerID:        assignment.JobID,
+		SizeBytes:      sizeBytes,
+		ChunkSizeBytes: fileArtifactChunkSize,
+		Checksum:       checksum,
+		IdempotencyKey: "file-read:" + assignment.JobID,
+	})
+	if err != nil {
+		return err
+	}
+	received := map[int]bool{}
+	for _, index := range opened.ReceivedChunkIndexes {
+		received[index] = true
+	}
+	file, err := os.Open(targetPath)
+	if err != nil {
+		return err
+	}
+	defer file.Close()
+	buffer := make([]byte, fileArtifactChunkSize)
+	for index, offset := 0, int64(0); offset < sizeBytes; index, offset = index+1, offset+int64(fileArtifactChunkSize) {
+		length := fileArtifactChunkSize
+		if remaining := sizeBytes - offset; remaining < int64(length) {
+			length = int(remaining)
+		}
+		if received[index] {
+			continue
+		}
+		if err := ctx.Err(); err != nil {
+			return err
+		}
+		read, err := file.ReadAt(buffer[:length], offset)
+		if err != nil && !(errors.Is(err, io.EOF) && read == length) {
+			return err
+		}
+		if read != length {
+			return fmt.Errorf("file artifact chunk is shorter than expected")
+		}
+		chunk := append([]byte(nil), buffer[:length]...)
+		state, err := worker.registeredState()
+		if err != nil {
+			return err
+		}
+		if _, err := worker.client.UploadArtifactChunk(ctx, protocol.ArtifactChunkUploadRequest{
+			RunEndpointID: state.RunEndpointID,
+			SessionToken:  state.SessionToken,
+			TransferID:    opened.TransferID,
+			ArtifactID:    artifactID,
+			ChunkIndex:    index,
+			Offset:        offset,
+			SizeBytes:     length,
+			Checksum:      bytesChecksum(chunk),
+			Payload:       chunk,
+		}); err != nil {
+			return err
+		}
+	}
+	state, err = worker.registeredState()
+	if err != nil {
+		return err
+	}
+	completed, err := worker.client.CompleteArtifactTransfer(ctx, protocol.ArtifactTransferCompleteRequest{
+		RunEndpointID: state.RunEndpointID,
+		SessionToken:  state.SessionToken,
+		TransferID:    opened.TransferID,
+		ArtifactID:    artifactID,
+		Checksum:      checksum,
+		SizeBytes:     sizeBytes,
+	})
+	if err != nil {
+		return err
+	}
+	if !completed.Completed || completed.Artifact.State != "available" {
+		return fmt.Errorf("artifact transfer did not complete")
+	}
+	return nil
+}
+
+func checksumServerFileForArtifact(ctx context.Context, targetPath string) (string, error) {
+	file, err := os.Open(targetPath)
+	if err != nil {
+		return "", err
+	}
+	defer file.Close()
+	hash := sha256.New()
+	buffer := make([]byte, fileArtifactChunkSize)
+	for {
+		if err := ctx.Err(); err != nil {
+			return "", err
+		}
+		read, err := file.Read(buffer)
+		if read > 0 {
+			if _, writeErr := hash.Write(buffer[:read]); writeErr != nil {
+				return "", writeErr
+			}
+		}
+		if errors.Is(err, io.EOF) {
+			break
+		}
+		if err != nil {
+			return "", err
+		}
+	}
+	return "sha256:" + hex.EncodeToString(hash.Sum(nil)), nil
+}

runtime/file_execution.go

@@ -101,6 +101,21 @@ func (executor *FileExecutor) scopeForAssignment(assignment protocol.RunJobAssig
 	return scope, false, err
 }
 
+func (executor *FileExecutor) existingReadTargetForAssignment(assignment protocol.RunJobAssignment) (string, string, bool, error) {
+	scope, deploymentRoot, err := executor.scopeForAssignment(assignment)
+	if err != nil {
+		return "", "", false, err
+	}
+	targetKey := assignment.TargetKey
+	if deploymentRoot {
+		targetKey = deploymentTargetKey(targetKey)
+		filePath, err := existingDeploymentTarget(scope, targetKey)
+		return scope, filePath, deploymentRoot, err
+	}
+	filePath, err := executor.resolver.ExistingTarget(scope, targetKey)
+	return scope, filePath, deploymentRoot, err
+}
+
 type fileListEntry struct {
 	Name         string `json:"name"`
 	Kind         string `json:"kind"`
@@ -175,11 +190,12 @@ func (executor *FileExecutor) list(ctx context.Context, scope string, deployment
 		if err != nil {
 			return err
 		}
-		logicalKey := path.Join(directoryKey, rel)
+		entryRelativePath := rel
 		if relativePath != "." {
-			logicalKey = path.Join(directoryKey, relativePath, rel)
+			entryRelativePath = path.Join(relativePath, rel)
 		}
-		candidate := append(entries, fileListEntry{Name: name, Kind: kind, RelativePath: rel, LogicalKey: logicalKey, SizeBytes: entryInfo.Size(), ModifiedAt: entryInfo.ModTime().UTC().Format(time.RFC3339Nano)})
+		logicalKey := path.Join(directoryKey, entryRelativePath)
+		candidate := append(entries, fileListEntry{Name: name, Kind: kind, RelativePath: entryRelativePath, LogicalKey: logicalKey, SizeBytes: entryInfo.Size(), ModifiedAt: entryInfo.ModTime().UTC().Format(time.RFC3339Nano)})
 		body, marshalErr := json.Marshal(fileListEnvelope{DirectoryKey: directoryKey, Path: strings.TrimPrefix(relativePath, "."), Entries: candidate})
 		if marshalErr != nil {
 			return marshalErr

runtime/worker.go

@@ -8,6 +8,7 @@ import (
 	"errors"
 	"fmt"
 	"log"
+	"net/http"
 	"os"
 	"path/filepath"
 	"runtime"
@@ -46,7 +47,10 @@ type WorkerClient interface {
 
 const (
 	jobActivePollInterval       = 10 * time.Second
+	jobClaimWaitSeconds         = 25
 	durableUploaderFlushTimeout = 5 * time.Second
+	fileArtifactChunkSize       = 1024 * 1024
+	maxFileArtifactBytes        = int64(512 * 1024 * 1024)
 )
 
 type Worker struct {
@@ -436,6 +440,7 @@ func (worker *Worker) ClaimAndRunOnce(ctx context.Context) (bool, error) {
 		SessionToken:  state.SessionToken,
 		Capabilities:  state.Capabilities,
 		Capacity:      worker.capacityReportFor(state),
+		WaitSeconds:   jobClaimWaitSeconds,
 	})
 	if err != nil {
 		if sessionInvalidError(err) {
@@ -728,7 +733,7 @@ func (worker *Worker) executeAssignment(ctx context.Context, assignment protocol
 	}
 	if supportedCapability(SupportedFileCapabilities(), assignment.Capability) {
 		log.Printf("RUN phase=job.dispatch status=selected job=%s executor=file", assignment.JobID)
-		return worker.executor.ExecuteContext(ctx, assignment)
+		return worker.executeFileJob(ctx, assignment)
 	}
 	if isSupportedDistributionCapability(assignment.Capability) {
 		log.Printf("RUN phase=job.dispatch status=selected job=%s executor=distribution", assignment.JobID)
@@ -742,6 +747,50 @@ func (worker *Worker) executeAssignment(ctx context.Context, assignment protocol
 	return lifecycleFailure("unsupported_run_capability", "unsupported run capability")
 }
 
+func (worker *Worker) executeFileJob(ctx context.Context, assignment protocol.RunJobAssignment) LifecycleExecutionResult {
+	if assignment.Capability != protocol.RunCapabilityFilesRead {
+		return worker.executor.ExecuteContext(ctx, assignment)
+	}
+	fileExecutor := worker.executor.fileExecutor
+	if fileExecutor == nil {
+		return lifecycleExecutionFailure("file_executor_unavailable", "file executor is unavailable", false)
+	}
+	scope, targetPath, deploymentRoot, err := fileExecutor.existingReadTargetForAssignment(assignment)
+	if err != nil {
+		return lifecycleExecutionFailure("file_read_failed", err.Error(), false)
+	}
+	info, err := os.Stat(targetPath)
+	if err != nil {
+		return lifecycleExecutionFailure("file_read_failed", err.Error(), false)
+	}
+	limit := assignment.ExecutionInput.MaxReadBytes
+	if limit <= 0 || limit > maxExecutionContentBytes {
+		limit = maxExecutionContentBytes
+	}
+	if info.Size() <= int64(limit) {
+		return fileExecutor.read(ctx, scope, deploymentRoot, assignment)
+	}
+	if info.Size() > maxFileArtifactBytes {
+		return lifecycleExecutionFailure("file_read_too_large", "file exceeds artifact transfer limit", false)
+	}
+	checksum, err := checksumServerFileForArtifact(ctx, targetPath)
+	if err != nil {
+		if errors.Is(err, context.Canceled) {
+			return lifecycleExecutionFailure("file_cancelled", "file read cancelled", false)
+		}
+		return lifecycleExecutionFailure("file_read_failed", err.Error(), false)
+	}
+	artifactID := "artifact-" + safeWorkspaceName(assignment.JobID) + "-file-read"
+	if err := worker.uploadFileArtifact(ctx, assignment, artifactID, targetPath, info.Size(), checksum); err != nil {
+		if errors.Is(err, context.Canceled) {
+			return lifecycleExecutionFailure("file_cancelled", "file read cancelled", false)
+		}
+		return lifecycleExecutionFailure("file_artifact_upload_failed", "file artifact upload failed", false)
+	}
+	metadata := fileExecutor.metadata(scope, assignment.TargetKey, checksum, info.Size())
+	return LifecycleExecutionResult{State: lifecycleResultStateSucceeded, Progress: protocol.RunJobProgressReport{Percent: 100, Message: "file read completed"}, ResultRef: "artifact://" + artifactID, Message: "file read completed", ExecutionResult: protocol.RunJobExecutionResult{Kind: "file.read", Version: metadata.Version, Checksum: checksum, SizeBytes: info.Size(), Summary: "large file transferred as artifact"}}
+}
+
 func (worker *Worker) executeProtectedRequestJob(ctx context.Context, assignment protocol.RunJobAssignment) LifecycleExecutionResult {
 	if err := protocol.ValidateRunJobAssignment(assignment); err != nil {
 		return protectedRequestFailure(assignment.Capability, ProtectedRequestStatusFailed, "protected_request_assignment_invalid")
@@ -988,31 +1037,38 @@ func (worker *Worker) reportRunUpdateHealth(ctx context.Context) error {
 }
 
 func (worker *Worker) runJobLoop(ctx context.Context, interval time.Duration) error {
-	log.Printf("RUN phase=job_loop status=starting pollMs=%d", interval.Milliseconds())
+	log.Printf("RUN phase=job_loop status=starting fallbackPollMs=%d claimWaitSeconds=%d", interval.Milliseconds(), jobClaimWaitSeconds)
-	ticker := time.NewTicker(interval)
-	defer ticker.Stop()
 	for {
 		select {
 		case <-ctx.Done():
 			log.Printf("RUN phase=job_loop status=context_done error=%s", RedactText(ctx.Err().Error()))
 			return ctx.Err()
-		case <-ticker.C:
+		default:
+		}
 		if worker.journal.ActiveCount() > 0 {
 			log.Printf("RUN phase=job_loop status=active_jobs activeJobs=%d", worker.journal.ActiveCount())
 			if err := worker.ReconcileOnce(ctx); err != nil {
 				log.Printf("RUN phase=job_loop status=reconcile_failed error=%s", RedactText(err.Error()))
-					ticker.Reset(boundedRetryBackoff(worker.cfg.RetryBackoff))
+				if err := waitWorkerLoop(ctx, boundedRetryBackoff(worker.cfg.RetryBackoff)); err != nil {
+					return err
+				}
 				continue
 			}
 			if err := worker.RecoverActiveJobs(ctx); err != nil {
 				log.Printf("RUN phase=job_loop status=recover_failed error=%s", RedactText(err.Error()))
-					ticker.Reset(boundedRetryBackoff(worker.cfg.RetryBackoff))
+				if err := waitWorkerLoop(ctx, boundedRetryBackoff(worker.cfg.RetryBackoff)); err != nil {
+					return err
+				}
 				continue
 			}
 		}
-			if _, err := worker.ClaimAndRunOnce(ctx); err != nil {
+		claimStartedAt := time.Now()
+		handled, err := worker.ClaimAndRunOnce(ctx)
+		if err != nil {
 			log.Printf("RUN phase=job_loop status=claim_failed error=%s", RedactText(err.Error()))
-				ticker.Reset(boundedRetryBackoff(worker.cfg.RetryBackoff))
+			if err := waitWorkerLoop(ctx, boundedRetryBackoff(worker.cfg.RetryBackoff)); err != nil {
+				return err
+			}
 			continue
 		}
 		worker.restartMu.Lock()
@@ -1022,9 +1078,26 @@ func (worker *Worker) runJobLoop(ctx context.Context, interval time.Duration) er
 			log.Printf("RUN phase=job_loop status=restart_requested")
 			return ErrSelfUpdateRestartRequested
 		}
-			ticker.Reset(interval)
+		if !handled && time.Since(claimStartedAt) < interval {
+			if err := waitWorkerLoop(ctx, interval-time.Since(claimStartedAt)); err != nil {
+				return err
 			}
 		}
+	}
+}
+
+func waitWorkerLoop(ctx context.Context, delay time.Duration) error {
+	if delay <= 0 {
+		return nil
+	}
+	timer := time.NewTimer(delay)
+	defer timer.Stop()
+	select {
+	case <-ctx.Done():
+		return ctx.Err()
+	case <-timer.C:
+		return nil
+	}
 }
 
 type durableLogClient interface {
@@ -1112,6 +1185,11 @@ func logBatchSessionMetadataMismatchError(err error) bool {
 	return errors.As(err, &mismatch) && mismatch.LogBatchSessionMetadataMismatch()
 }
 
+func artifactChunkMissingOnPlatform(err error) bool {
+	var httpErr interface{ HTTPStatus() int }
+	return errors.As(err, &httpErr) && httpErr.HTTPStatus() == http.StatusNotFound
+}
+
 type sessionArtifactChunkClient struct {
 	client        durableArtifactClient
 	runEndpointID string
@@ -1121,7 +1199,15 @@ type sessionArtifactChunkClient struct {
 func (client sessionArtifactChunkClient) UploadArtifactChunk(ctx context.Context, chunk protocol.ArtifactChunkUploadRequest) (protocol.ArtifactChunkUploadResponse, error) {
 	chunk.RunEndpointID = client.runEndpointID
 	chunk.SessionToken = client.sessionToken
-	return client.client.UploadArtifactChunk(ctx, chunk)
+	response, err := client.client.UploadArtifactChunk(ctx, chunk)
+	if err != nil {
+		if artifactChunkMissingOnPlatform(err) {
+			log.Printf("RUN phase=durable_uploaders.artifacts status=drop_stale transfer=%s artifact=%s chunk=%d reason=platform_transfer_missing", safeOptional(chunk.TransferID), safeOptional(chunk.ArtifactID), chunk.ChunkIndex)
+			return protocol.ArtifactChunkUploadResponse{Accepted: true, TransferID: chunk.TransferID, ArtifactID: chunk.ArtifactID, ChunkIndex: chunk.ChunkIndex}, nil
+		}
+		return response, err
+	}
+	return response, nil
 }
 
 func (worker *Worker) runDurableUploaders(ctx context.Context, done chan<- struct{}) {

runtime/worker_test.go

@@ -7,6 +7,8 @@ import (
 	"fmt"
 	"net/http"
 	"net/http/httptest"
+	"os"
+	"path/filepath"
 	"reflect"
 	"runtime"
 	"strings"
@@ -173,6 +175,9 @@ func TestWorkerClaimsAcksProgressAndCompletesJob(t *testing.T) {
 	if len(client.ackRequests) != 1 || len(client.progressRequests) != 1 || len(client.resultRequests) != 1 || len(client.cancelPollRequests) != 1 {
 		t.Fatalf("expected ack/progress/result/cancel calls, got ack=%d progress=%d result=%d cancel=%d", len(client.ackRequests), len(client.progressRequests), len(client.resultRequests), len(client.cancelPollRequests))
 	}
+	if len(client.claimRequests) != 1 || client.claimRequests[0].WaitSeconds != jobClaimWaitSeconds {
+		t.Fatalf("expected persistent claim wait, got %+v", client.claimRequests)
+	}
 	if client.progressRequests[0].Progress.Percent != 10 || client.resultRequests[0].State != "succeeded" || client.resultRequests[0].ResultRef == "" {
 		t.Fatalf("unexpected job channel payloads: progress=%+v result=%+v", client.progressRequests[0], client.resultRequests[0])
 	}
@@ -214,6 +219,56 @@ func TestWorkerDispatchesSelfUpdateJob(t *testing.T) {
 	}
 }
 
+func TestWorkerFileReadLargeFileUploadsArtifact(t *testing.T) {
+	cfg := workerTestConfig(t)
+	client := newFakeWorkerClient()
+	assignment := workerJobAssignment(protocol.RunCapabilityFilesRead)
+	assignment.TargetKey = "logs/big.log"
+	assignment.ExecutionInput.WorkspaceScope = "run-local"
+	assignment.ExecutionInput.MaxReadBytes = 4
+	client.claimJob = assignment
+	client.buildInput = protocol.DistributionBuildInputResponse{JobID: assignment.JobID}
+	scope, err := NewWorkspaceResolver(cfg.WorkspaceRoot).Scope(assignment.ServerInstanceID, assignment.ExecutionInput.WorkspaceScope)
+	if err != nil {
+		t.Fatalf("scope: %v", err)
+	}
+	if err := os.MkdirAll(filepath.Join(scope, "logs"), 0o700); err != nil {
+		t.Fatalf("mkdir fixture: %v", err)
+	}
+	payload := []byte(strings.Repeat("A", fileArtifactChunkSize) + "tail")
+	if err := os.WriteFile(filepath.Join(scope, assignment.TargetKey), payload, 0o600); err != nil {
+		t.Fatalf("write fixture: %v", err)
+	}
+	worker, err := NewWorker(cfg, client)
+	if err != nil {
+		t.Fatalf("new worker: %v", err)
+	}
+	if err := worker.Register(context.Background()); err != nil {
+		t.Fatalf("register: %v", err)
+	}
+
+	handled, err := worker.ClaimAndRunOnce(context.Background())
+	if err != nil || !handled {
+		t.Fatalf("claim/run handled=%v err=%v", handled, err)
+	}
+	if len(client.artifactOpenRequests) != 1 {
+		t.Fatalf("expected artifact transfer, got %+v", client.artifactOpenRequests)
+	}
+	opened := client.artifactOpenRequests[0]
+	if opened.OwnerID != assignment.JobID || opened.SizeBytes != int64(len(payload)) || opened.Checksum != bytesChecksum(payload) || opened.ChunkSizeBytes != fileArtifactChunkSize {
+		t.Fatalf("unexpected artifact open: %+v", opened)
+	}
+	if string(client.artifactPayload) != string(payload) {
+		t.Fatalf("uploaded artifact payload mismatch")
+	}
+	if len(client.resultRequests) != 1 || client.resultRequests[0].State != "succeeded" || !strings.HasPrefix(client.resultRequests[0].ResultRef, "artifact://artifact-job-worker-file-read") {
+		t.Fatalf("expected artifact result ref, got %+v", client.resultRequests)
+	}
+	if client.resultRequests[0].ExecutionResult.Content != "" || client.resultRequests[0].ExecutionResult.SizeBytes != int64(len(payload)) {
+		t.Fatalf("large file result must not inline content: %+v", client.resultRequests[0].ExecutionResult)
+	}
+}
+
 func TestWorkerRegistersPackageIdentity(t *testing.T) {
 	client := newFakeWorkerClient()
 	cfg := workerTestConfig(t)
@@ -608,6 +663,19 @@ func TestSessionArtifactChunkClientOverridesSpooledIdentity(t *testing.T) {
 	}
 }
 
+func TestSessionArtifactChunkClientDropsPlatformMissingTransfer(t *testing.T) {
+	recorder := &recordingDurableArtifactClient{err: api.PlatformRequestError{Status: http.StatusNotFound, Code: "not_found"}}
+	chunk := protocol.ArtifactChunkUploadRequest{RunEndpointID: "old-endpoint", SessionToken: "old-token", TransferID: "transfer-stale", ArtifactID: "artifact-stale", ChunkIndex: 3}
+	client := sessionArtifactChunkClient{client: recorder, runEndpointID: "run-current", sessionToken: "token-current"}
+	response, err := client.UploadArtifactChunk(context.Background(), chunk)
+	if err != nil || !response.Accepted || response.TransferID != chunk.TransferID || response.ChunkIndex != chunk.ChunkIndex {
+		t.Fatalf("expected stale chunk ack for queue cleanup, response=%+v err=%v", response, err)
+	}
+	if recorder.chunk.RunEndpointID != "run-current" || recorder.chunk.SessionToken != "token-current" {
+		t.Fatalf("expected current identity before stale drop, got %+v", recorder.chunk)
+	}
+}
+
 func TestWorkerRetryBackoffIsBounded(t *testing.T) {
 	if got := boundedRetryBackoff(75 * time.Millisecond); got != 75*time.Millisecond {
 		t.Fatalf("expected configured backoff, got %s", got)
@@ -640,7 +708,7 @@ func TestWorkerIntegrationWithPlatformLikeServer(t *testing.T) {
 		case "/api/v1/run/jobs/claim":
 			var request protocol.RunJobClaimRequest
 			decodeWorkerTestJSON(t, r, &request)
-			if request.SessionToken != "session-token" || len(request.Capabilities) == 0 {
+			if request.SessionToken != "session-token" || len(request.Capabilities) == 0 || request.WaitSeconds != jobClaimWaitSeconds {
 				t.Fatalf("unexpected claim: %+v", request)
 			}
 			writeWorkerTestJSON(t, w, protocol.RunJobClaimResponse{Accepted: true, RunEndpointID: request.RunEndpointID, HasJob: true, Job: &assignment, NextPollSeconds: 2, ServerTime: workerTestTime()})
@@ -811,10 +879,14 @@ func (client *recordingDurableLogClient) IngestLogBatch(_ context.Context, batch
 
 type recordingDurableArtifactClient struct {
 	chunk protocol.ArtifactChunkUploadRequest
+	err   error
 }
 
 func (client *recordingDurableArtifactClient) UploadArtifactChunk(_ context.Context, chunk protocol.ArtifactChunkUploadRequest) (protocol.ArtifactChunkUploadResponse, error) {
 	client.chunk = chunk
+	if client.err != nil {
+		return protocol.ArtifactChunkUploadResponse{}, client.err
+	}
 	return protocol.ArtifactChunkUploadResponse{Accepted: true, TransferID: chunk.TransferID, ArtifactID: chunk.ArtifactID, ChunkIndex: chunk.ChunkIndex}, nil
 }