# Delivery comment template

Post this comment on the source ticket once the recovered production has actually published,
not when the render finishes. Every checkbox should point at something you can verify, not
something you assume happened.

```markdown
**Recovery result: published and verified.**

- [ ] Quality gate: cleared on <date>, after the output passed the master QC checklist
      (link the checklist run or the comment recording it).
- [ ] Publish step: ran on <date> at <time>, target: <platform / channel>.
- [ ] Public verification: the published item is reachable and shows the expected content
      (link the public URL, and the HTTP status or screenshot used to confirm it).
- [ ] Callback: a confirmation was sent back to this ticket automatically
      (link the callback comment or activity entry, not just "it should have run").
- [ ] Retry scope closed: every step listed in the recovery ticket's retry scope reached Done
      (link the recovery ticket).

**What was preserved:** <count of steps left untouched, and why they did not need to be
retried>.

**Known limits:** <anything checked and found not fully resolved, e.g. a workaround instead of
a code fix, so the next person does not assume the underlying cause is gone>.
```

## Real example (redacted)

```markdown
**Recovery result: published and verified.**

- [x] Quality gate: cleared 23 August, after 1920×1080 / 755.167 s video vs 755.160 s audio /
      -19.8 LUFS / no audio silence over 2 s passed the master QC checklist.
- [x] Publish step: ran 23 August at 18:15, target: YouTube (public).
- [x] Public verification: https://www.youtube.com/watch?v=slaxpdQFlo4 returns HTTP 200 and
      shows the expected title.
- [x] Callback: confirmation posted automatically to the source production ticket.
- [x] Retry scope closed: all 18 clips and 11 images from the recovery ticket reached Done.

**What was preserved:** 89 clips and 107 images already correct, left untouched throughout
the recovery.

**Known limits:** the final assembly step's repeated six-minute failure was worked around by
re-running it after conditions changed. No dedicated code fix for that failure was found in
the available record; track it as an open risk on the next long-form production.
```