user-engine/workplans/USER-WP-0032-platform-service-operations.md

---
id: USER-WP-0032
type: workplan
title: "P05 checked service health and bounded delivery recovery"
domain: communication
repo: user-engine
status: finished
owner: codex
topic_slug: communication
created: "2026-09-13"
updated: "2026-09-13"
state_hub_workstream_id: "35af93fc-17ed-548e-8da7-616e44e8eee0"
---

Complete P05 under USER-WP-0030-T03. P06 remains separate. Do not send production
messages for acceptance; use the existing loopback SMTP/IMAP harness. Live SMTP
authentication diagnostics do not submit a message or claim inbox delivery.

## Expose checked service state with clear recovery steps

```task
id: USER-WP-0032-T01
status: done
priority: high
state_hub_task_id: "0c9ebc81-ea7c-597e-a908-2806e5a8c57f"
```

Operator-only, bounded checks for KeyCape dependencies/current factor-read
credential and mail database/SMTP authentication. Show checked time, impact,
unknown versus unavailable state, retry and assisted setup; no secrets or raw
provider errors. Health probes must not send mail.

## Make selected delivery and retry safe and reviewable

```task
id: USER-WP-0032-T02
status: done
priority: high
state_hub_task_id: "6be221b9-5378-54a9-ae12-37a52cc9e186"
```

Fix failed worker claims; serialize event operations across connections; confirm
one selected delivery and reject stale/repeated completion. Missing email lane
must not mark invitations accepted. Preserve provider idempotency across lost
responses and restart; uncertain SMTP outcomes must not auto-resend. No automatic
processing of the production backlog during this task.

## Verify failure and recovery through native and harness acceptance

```task
id: USER-WP-0032-T03
status: done
priority: high
state_hub_task_id: "5ee8bb39-e725-592e-baf2-04063c579ccd"
```

Installed-provider genuine JWT expiry, scope withdrawal/recovery and outage;
native renewal/projection; SMTP/IMAP received-message and retry tests with a
local mailbox; role/CSRF/stale/ambiguous-outcome browser coverage. Publish digest
pinned releases and verify live non-sending status. Close RPF-WP-0040 acceptance
using its specified isolated fixture rather than disrupting live authentication.


Completed 2026-09-13. Evidence: docs/evidence/2026-09-13-p05-service-operations.md.
227 portal tests with disposable PostgreSQL, 23 browser checks, 89 email tests,
real local SMTP/IMAP one-message retry reconciliation, installed-provider expiry /
permission recovery and native credential rotation passed. Three exact-commit
CI image releases are deployed and all six live component checks pass without
sending production mail. The invitation URL and historical Never-pull drift
are corrected in live and canonical configuration.

Residuals already have live owners before closure: P06 is USER-WP-0030-T03 /
KEY-WP-0035; actual user inbox/acceptance journeys remain USER-WP-0028, and the
historical resolver incident remains NK-WP-0033. P05 does not claim these done.