The original incident page says one issue and one queued fix. The exported implementation distinguishes a shared-parser fault from a booking-form hint mismatch. Similar subjects and similar date swaps do not make them the same fault.
What the exported data shows
Counts below are computed by the kb component from its report data. It deduplicates by report ID before selecting reports merged into ticket 4471. Repeated export rows are not additional people or additional reports.
412 reports merged into #4471 · 2 causes
| cause | reports | tagged as | tenants | seen | |
|---|---|---|---|---|---|
parser | 379 | parser ×356 · hint ×23 | 11 | 2026-04-01 → 2026-10-04 | #4471 |
hint | 33 | hint ×18 · parser ×15 | 3 | 2026-04-17 → 2026-09-22 | not this ticket — no ticket of its own |
5 reports read as a one-day move; filed under parser by tenant.
The table's ticket-ownership labels describe the original support export, not GitHub issues. The two problems are tracked separately in the shared-parser issue and the booking-v3 hint issue.
Shared parser
The shared parser reads month-first while these tenants write dates day-first. For example, R-41200 from harbour.dental reports a reminder for 7 April arriving for 4 July. The tenant directory identifies its locale as en-GB.
The exported classifier assigns day-first tenant reports here except form-channel reports using booking-v3. Use the tenant's stored locale, not the reporter's browser locale or intake tag. Fix and test parsing separately from the displayed form hint.
Booking-v3 date hint
Booking-v3 displays MM/DD/YYYY, but its form reads the date using the tenant's
convention. For day-first tenants, following that hint can swap the intended date.
R-41230 from copper.studio reports 1 June instead of 6 January; the tenant uses
es-ES and booking-v3, and the report channel is form.
This group is limited to form-channel reports for day-first booking-v3 tenants. Their app and email reports stay in the parser group. Align the hint with the accepted input and test both together; a parser fix alone does not demonstrate this mismatch is resolved.
Keep uncertain reports visible
The desk notes flag R-41202, R-41209, R-41236, R-41344 and R-41620 as unusual timing or one-day moves. The exported classifier assigns them to parser, but their wording does not independently establish that cause. Keep their IDs counted exactly once and investigate them; do not turn a possible timezone explanation into a confirmed bug.
Triage and follow-up
The written triage rule and accepted decision 0001 merge on normalized subject alone. The interactive example instead compares causes. This disagreement explains why the written policy is not enough to establish that every merged report shares a fix.
Proposed practice: compare the report body and the tenant's locale, form and channel before merging. Matching subjects are a search aid, not proof of a shared cause. This is a recommendation, not an approved policy change. Leave accepted decision 0001 unchanged; any replacement needs a new decision record.
The grouping evidence retains source hashes and per-report assignments for audit. The new download did not establish whether the desk has received additional reports outside this export. The counts above apply only to the supplied snapshot.
desk ref DR-AF0J-5Y3W