Data cleanup: 7 live events with malformed/incomplete dates (`2026-07-??`, `2026-??-??`) — 1 published, 6 draft

[chubes4]

Summary

Investigation into the 2026-07-?? fatal (PR #393, issue #394) surfaced 7 existing events on events.extrachill.com (blog 7) whose event-details block stores an incomplete/malformed date. One is published and was the live fatal trigger; six are drafts. All 7 have a coerced 0000-00-00 00:00:00 row in the c8c_7_datamachine_event_dates custom table (MySQL silently zeroed the malformed string under non-strict sql_mode).

Affected events

ID	Title	Status	Stored startDate / endDate
261090	Preservation All-Stars	publish	2026-07-?? / 2026-07-??
222758	Secret Planet Presents: Puuluup	draft	2026-09-?? / 2026-09-??
224284	MAE [NIGHT 1]	draft	2026-04-?? / 2026-04-??
253597	Hotel Sewing Kit	draft	2026-??-?? / 2026-??-?? (year only)
256872	DERYA YILDIRIM & GRUP ŞİMŞEK	draft	2026-07-?? / 2026-07-??
257930	PLAYBOY MANBABY: Corporations Are People Tour	draft	2026-07-?? / 2026-07-??
258783	PREWN	draft	2026-07-?? / 2026-07-??

All 7 custom-table rows: start_datetime = end_datetime = 0000-00-00 00:00:00.

Why this happened

These are real "announced, exact day TBA" listings (e.g. Preservation Hall's "July 2026"). The AI extraction emitted the incomplete date literally and nothing validated it before publish/storage. Root cause + refactor tracked in #394.

Cleanup actions

• 261090 (published) — immediate: un-publish (→ draft) so it's not a landmine for any other date-consuming path, OR resolve the real date if it can be found on the venue site and republish with a valid Y-m-d.
• The 6 drafts — decide per the Stabilize event date handling: validate at the AI gate + storage chokepoint, route render through a non-throwing factory #394 product question (reject/delete vs. hold as "date TBA"). Since they're already unpublished, no urgency, but they should not be left to auto-resolve into bad published rows.
• Once Stabilize event date handling: validate at the AI gate + storage chokepoint, route render through a non-throwing factory #394 lands a validator, re-scan blog 7 for any other REGEXP '20[0-9][0-9]-[0-9?][0-9?]-[?]' rows and the zeroed 0000-00-00 table rows, and reconcile.

Detection query (for re-scans)

SELECT ID, post_title, post_status FROM c8c_7_posts
WHERE post_content REGEXP '20[0-9][0-9]-[0-9?][0-9?]-[?]';

SELECT post_id FROM c8c_7_datamachine_event_dates
WHERE start_datetime = '0000-00-00 00:00:00';

Note

Blocked-by / related-to #394 for the systemic fix. This issue is the data remediation only.

[chubes4]

Immediate safety action taken: 261090 (Preservation All-Stars) un-published → draft. The real date is not resolvable from the venue site (preservationhall.com lists it behind the ticketing calendar with no concrete day — consistent with a genuine 'July 2026, day TBA' listing), so unpublishing was the right call rather than guessing a Y-m-d.

Verified: zero published events with malformed dates remain on blog 7. The 6 drafts are left in place pending the #394 product decision (reject vs. TBA-state vs. hold-as-draft).