Field reportDT-2026-0181
Embedding drift is a migration problem, not a model problem
Changing the embedding model invalidates the index. Few teams plan for it.
7 minBig Data & Vector DBs
Vectors from two different models are not comparable. Upgrading means re-embedding the whole corpus, which for large stores is a multi-day job with a consistency window in the middle.
Teams that version the index and run both during cutover treat this as routine. Teams that do not discover it during an incident.