Repair the source attendance data before trusting the counts

Mirta, 2026-10-08: 'hay varios problemas con el documento, no esta actualizado tampoco con el ultimo webinario, entiendo que primero habria que arreglar el documento antes de que intente ejecutar nada pq sino va a tirar todo mal.' She is right. Under 3/6/9/12, 30 of 382 migrated sends are one short of their threshold. Three separate repairs, only the first is the app's job: 1. Webinars after her 2026-09-10 export. Newest webinar in the DB is 2026-08-05 and the series runs roughly monthly, so one or two sessions are missing outright. Fix: she exports each attendee report and uploads it through /ingest. No spreadsheet work. 2. 2025-10-22, the #REF! session: 10 of 15 hand-added addresses destroyed in her file, unrecoverable from it. Needs the original attendee report re-exported and ingested. Until then Kutloano Moshao and others stay one webinar short of their tier. 3. The other thin tabs: migration/README.md line 106 - 8 of 77 tabs have fewer distinct emails than their stated Unique Viewers (six by exactly one, 23.12.13 by -10), one has more. Each needs its own re-export. Blocks: the 'print the derived count or hers' question (she declined to answer until the data is fixed), and confidence in any pending list. Also unconfirmed: nushjaleel@gmail.com SuperStar has sent_on 2015-08-13, obvious typo for 2025.