Reduce pending input scan interval to one second

CONTRIBUTING.md

@@ -1175,16 +1175,18 @@ The connection owner spans preparation and the transferred Run without a reserva
 deadline; every exit releases it.
 
 The existing Worker scans pending inputs using the same bounded scheduling slots,
-Session locks, durable deadlines and engine capability checks. Keep the five-second
+Session locks, durable deadlines and engine capability checks. Keep the one-second
 Environment-input scan cadence and at most 100 candidates per scan. At EOF after a
 nonempty cursor, refill the first page once in the same scan; an empty queue must
 not spin. Advance the cursor before readiness checks so an unavailable Runtime
 cannot starve later candidates. Preserve the four active slots and alternation
 between ordinary Turns and Environment inputs. A self-hosted Session
 waits for its dedicated enrolled device; it cannot select an arbitrary same-tenant
 device or migrate an existing binding. Preparation failure can retry while still
-pending without extending the deadline. Unknown promotion results or errors after
-admission retain the existing no-replay settlement rules.
+pending without extending the deadline. Preparation retries share this scan cadence;
+the four slots bound concurrency, not attempt frequency. Managed-provider lifecycle
+polling retains its separate five-second interval. Unknown promotion results or
+errors after admission retain the existing no-replay settlement rules.
 
 `AGENTS_API_DAEMON_WS_URL` enables the private gateway and supplies the unchanged
 public `remote_url`. `AGENTS_API_HARNESSES` explicitly adds deployment-supported

services/agents-api/internal/execution/worker_schedule.go

@@ -34,7 +34,7 @@ func (s *workerSchedule) selectWork(ctx context.Context, w *Worker, devices []st
 		if err != nil {
 			return nil, err
 		}
-		s.nextEnvironmentScan = time.Now().Add(5 * time.Second)
+		s.nextEnvironmentScan = time.Now().Add(time.Second)
 		if len(environments) == 0 && s.environmentCursor != "" {
 			s.environmentCursor = ""
 			// Retry the first page now instead of spending a scan interval on EOF.

services/agents-api/internal/store/environment_worker_scan_test.go

@@ -33,7 +33,7 @@ func TestWorkerEnvironmentRetriesNewlyReadyAtNextScan(t *testing.T) {
 	waitTurn(t, h, barrier.ID, store.TurnCompleted)
 
 	prepare := nextWorkerFrame(t, frames, proto.TypeExecutionPrepare)
-	if elapsed := time.Since(scanned); elapsed < 4*time.Second || elapsed > 8*time.Second {
+	if elapsed := time.Since(scanned); elapsed < 750*time.Millisecond || elapsed > 3*time.Second {
 		t.Fatal("readiness retry must use the next scan, without an empty scan interval", elapsed)
 	}
 	if workerRuntimeForPreparation(t, h, prepare) != runtime {
@@ -79,7 +79,7 @@ func TestWorkerEnvironmentPaginationReachesReadyTail(t *testing.T) {
 	// The first 100 unbound inputs must not pin the cursor, and the ready
 	// tail must wait for its own bounded page rather than an unbounded drain.
 	prepare := nextWorkerFrame(t, frames, proto.TypeExecutionPrepare)
-	if elapsed := time.Since(scanned); elapsed < 4*time.Second || elapsed > 8*time.Second {
+	if elapsed := time.Since(scanned); elapsed < 750*time.Millisecond || elapsed > 3*time.Second {
 		t.Fatal("pagination lost the scan bound or starved the ready tail", elapsed)
 	}
 	if workerRuntimeForPreparation(t, h, prepare) != runtime {

services/agents-api/internal/store/environment_worker_test.go

@@ -147,7 +147,7 @@ func TestWorkerEnvironmentRetriesPendingWithoutExtendingDeadline(t *testing.T) {
 	h.write(request.ID, proto.TypeDone, proto.DonePayload{Content: "complete"})
 	waitTurn(t, h, request.ID, store.TurnCompleted)
 	second := nextWorkerFrame(t, frames, proto.TypeExecutionPrepare)
-	if elapsed := time.Since(started); elapsed < 4*time.Second || elapsed > 8*time.Second || first.ID == second.ID {
+	if elapsed := time.Since(started); elapsed < 750*time.Millisecond || elapsed > 3*time.Second || first.ID == second.ID {
 		t.Fatal("pending preparation missed its next scan or reused a released owner")
 	}
 	acknowledgePreparation(runtime, second.ID)