Add migration to reconcile stale plot_count data

[realproject7]

Summary

• Add supabase/migrations/00013_reconcile_plot_counts.sql to fix existing stale storylines.plot_count and last_plot_time values
• Idempotent UPDATE using COUNT(*) and MAX(block_timestamp) from plots — only updates rows where values differ
• Companion to PR Fix plot_count: reconcile storyline after plot upsert #258 which fixed the code path going forward

Fixes #259

Test plan

• Migration updates stale storyline rows with correct plot_count and last_plot_time
• Running migration twice produces no additional changes (idempotent)
• npm run typecheck passes

🤖 Generated with Claude Code

[project7-interns]

REQUEST CHANGES — one blocking issue.

Blocking: != doesn't match NULL values — stale rows with NULL last_plot_time will be skipped.

Line 16: s.plot_count != sub.cnt OR s.last_plot_time != sub.latest

In PostgreSQL, NULL != value evaluates to NULL (not TRUE), so any storyline where last_plot_time is currently NULL (never set) won't be updated by this migration — which is likely the most common stale case.

Fix: use IS DISTINCT FROM instead of !=:

WHERE s.storyline_id = sub.storyline_id
  AND (s.plot_count IS DISTINCT FROM sub.cnt
    OR s.last_plot_time IS DISTINCT FROM sub.latest);

IS DISTINCT FROM treats NULL as a comparable value and returns TRUE when one side is NULL and the other isn't. Same applies to plot_count if it could be NULL.

Otherwise the SQL structure is correct — idempotent UPDATE with computed aggregates, only touches differing rows.

[project7-interns]

Verdict: REQUEST CHANGES

Summary

The migration is close, but the WHERE predicate uses SQL != comparisons that do not handle NULL correctly.

Findings

• [medium] Rows with NULL in storylines.last_plot_time can be skipped even when they need repair. In PostgreSQL, NULL != sub.latest evaluates to NULL, not TRUE, so the whole AND (...) condition may fail. Example: if a storyline already has the correct plot_count but a stale NULL last_plot_time, this migration will not update it.
  • File: supabase/migrations/00013_reconcile_plot_counts.sql:16
  • Suggestion: use IS DISTINCT FROM for both comparisons: s.plot_count IS DISTINCT FROM sub.cnt OR s.last_plot_time IS DISTINCT FROM sub.latest.

Decision

Request changes. The migration should use null-safe comparison so it reliably repairs all stale rows.

[project7-interns]

APPROVE

NULL-safety fix confirmed: IS DISTINCT FROM on both comparisons. Migration is now correct and idempotent — will properly catch stale rows regardless of NULL state.

[project7-interns]

Verdict: APPROVE

Summary

The follow-up patch fixes the migration predicate by using null-safe comparisons. The repair now correctly updates stale last_plot_time rows even when existing values are NULL.

Findings

• [info] No blocking issues found in the updated patch.

Decision

Approve. IS DISTINCT FROM makes the migration reliably idempotent and null-safe. CI was still pending when reviewed.