Documentation

Drives the connections store through the full cutover lifecycle (cassandra -> migration-to-postgresql -> postgresql) on a dynamic backend. Every phase creates a fresh pair of users and connects them via the brig public API, exercising the writes of all three interpreters (Cassandra, dual-write, Postgres). Connections written by earlier interpreters must remain readable afterwards: the pair connected during the pure-Cassandra phase proves the background worker backfilled rows that only ever existed in Cassandra.

Remote (connection_remote) rows are not seeded explicitly here; the migration's copy loop still runs over whatever dual-write produced, and federated connections are covered by the brig integration suite.