MARS-84 UX-pain digest can never reach its pattern threshold for parameterised toasts: feedbackKey embeds the entity id, so one problem fragments into 11 keys

The 2026-09-28 13:00Z digest (eventId 14124118-8ce5-440d-b080-7cda7eae94df) reported '62 avisos · 0 patrones (>=10) · 0 críticos (>=100)'. Its arithmetic is right and its conclusion is wrong: a single UX problem hit 19 times across 7 DISTINCT USERS in that window and was reported as no pattern at all. MEASURED, 7 days to 2026-09-28, action='error-shown': the family detail->>'feedbackKey' LIKE 'toast:cada-conducto%' totals 19 events / 7 users, but splits into ~11 keys because the PIEZA NUMBER IS PART OF THE KEY - toast:cada-conducto-de-la-pieza-36 (4), -21 (3), -44 (3), -12 (2), -13 (2), -14 (2), -23 (1), -25 (1), -27 (1), and more in the tail. Every one is below 10, so the >=10 gate cannot fire no matter how bad the problem gets. THE MECHANISM IS GENERAL, not one bad key. These feedbackKeys are slugified, truncated Spanish MESSAGE TEXT ('toast:ingresa-una-calificacion-entre-0-y', 'toast:selecciona-el-sexo-del-paciente'), so ANY message that interpolates a tooth, a field or a name fragments by construction, and the more users hit it across more entities the more it fragments. The threshold is hardest to reach exactly when the problem is most widespread - the failure mode is inverted. SECOND, SMALLER DEFECT on the same rows: detail->>'code' is NULL on all 62 events. If any aggregation keys on 'code' it groups everything into one bucket; the usable discriminator is feedbackKey. Suggested direction, not a decision: derive the pattern key from a STABLE message template id rather than the rendered string, or strip the interpolated segment before grouping. A digest whose threshold is unreachable is worse than no digest - it publishes a weekly all-clear. Coverage: I measured the 'error-shown' action only, on mars only, over 7 days. Whether nudge-shown/warn-tier keys fragment the same way is untested, and the same feedbackKey scheme on pluto/venus/terra/enamel is unexamined.