[Bug] Pagination defaults inconsistent + unbounded countDocuments + cron uses env not Settings

[chronoai-shining]

Severity: Medium

Three related API-surface hygiene issues.

1. Pagination defaults vary

Endpoint	Default pageSize	Max
/audit (admin/controllers/AuditController.ts:91-101)	50	200
/ingest/history (ingest/controllers/IngestController.ts:169-173)	20	100
/schemas (schemas/controllers/SchemaController.ts:108-112)	20	100
/workflows (workflows/controllers/WorkflowController.ts:97-101)	20	100
/connectors (workflows/controllers/ConnectorController.ts:102-106)	20	100
/runs/history (runner/controllers/HistoryController.ts:36-44)	20	100

No lower-bound clamp: pageSize=0 or negative bypasses pagination logic.

2. countDocuments(filter) on every list, no upper bound

Every list endpoint runs countDocuments(filter) in parallel with the find. At scale (100k+ docs) this is slower than the find itself. ConnectorController.syncConnectors (workflows/controllers/ConnectorController.ts:272-285) does find().toArray() with no limit — loads the entire catalog into memory.

3. Cron uses env, not Settings

runner/session-manager.ts:24-27 reads VERIFY_CRON_INTERVAL_HOURS once at module load. The Settings document (admin/models/settings.ts:11) stores verifyCronIntervalHours and lets operators update it via PUT /settings. Updates are silent no-ops until pod restart.

Same for Settings.eventRetentionDays — wired into the TTL index at boot, never reapplied.

Remediation

• Add a parsePagination(page, pageSize) util in shared/crud.ts (depends on [Feature] Centralize HTTP client + CRUD scaffolding (kill duplication) #33). Clamp 1 <= pageSize <= 100, page >= 1. Apply uniformly.
• Use estimatedDocumentCount() when no filter is applied; for filtered queries, allow a ?countMode=skip|estimate|exact query param so callers can opt out.
• For syncConnectors, paginate the upload batch or stream.
• Re-read Settings.verifyCronIntervalHours on boot; subscribe to settings changes (Mongo change stream OR a periodic poll on a settingsVersion field) and call stopVerifyCron(); startVerifyCron();. For expireAfterSeconds on event_logs, run db.runCommand({collMod: ...}) on settings change.

Coupled with #32 (config) and #33 (CRUD helpers).