I get "Index … is locked by pid", how do you clear it?

That error comes from a stale lock in the index_indexer_state table after a reindex was interrupted (deploy, killed process, OOM). We confirm no reindex is actually running, then clear the lock for that indexer and rerun the reindex cleanly. We never just delete locks blindly, we first verify nothing is mid-write so we do not corrupt the index.