A migration job was still running long after the release had finished. A read-only investigation found a lookup comparing a numeric key as text. The query plan showed a scan where an existing index could have been used. That gave the team a specific cause to investigate instead of another general report that the migration was slow.
The fix was small, but its handoff included more than the changed line. The Duva update linked the pull request, described the old and new query plans, and recorded a PostgreSQL regression test that failed with the original expression. A later development check confirmed the indexed lookup, while production rollout and acceptance remained separate work.
Carry the diagnosis with the fix. For a performance debugging handoff, include the observed symptom, the smallest useful evidence, the proposed change, and the check that would prove it helped. A benchmark for one query fragment should stay labeled as that fragment, rather than becoming a claim about the whole migration.