Indexers Are Databases, Treat Them Like It

Teams that would never run a production Postgres instance without migrations, backups, and a rollback plan will happily ship an indexer with none of the three.

Reorgs are just concurrent writes

A chain reorg is a write conflict. If your indexer does not have an explicit strategy for it — a confirmation depth, a rollback log, an idempotent write path — you have a database with untested concurrency handling in production.

Plan the backfill before you need it

The migration nobody plans for is the one that requires re-indexing from genesis. Build the backfill path on day one, even if you never expect to use it before you actually need it at 2am.