[rickyrombo]

The genesis migration no longer copies writes onto the new chain from the API. A standalone mirror in go-openaudio (openaudio mirror, OpenAudio/go-openaudio#605, #608, #610) now replays the old chain's ManageEntity and Plays transactions onto the new chain in block order. The API flusher is already off in production, so the queue, the flusher and their config are dead.

Removed

• api/new_chain_flusher.go and api/new_chain_flusher_test.go.
• Construction and startup of the flusher in api/server.go.
• The only insert into new_chain_queue: the fire-and-forget enqueueForNewChain call after a confirmed relay in api/relay.go. The relay still submits to the chain and returns the same response.
• The new_chain_queue table. Migration 0243_drop_new_chain_queue.sql drops it. Its objects are removed from sql/01_schema.sql, and its model is removed from api/dbv1/models.go. I edited both by hand so a full regeneration's unrelated drift stays out of this PR.
• NewChainID and its per-environment defaults. The flusher was the only code that stamped it (fix(api): stamp the new chain ID on flushed transactions #1059).
• Config fields and env vars: newChainUrl, newChainId, newChainQueueEnabled, newChainFlushEnabled, newChainFlushFromBlock, newChainFlushToBlock, newChainInsecureSkipVerify.

Kept

• The go-openaudio pin from chore(deps): pin go-openaudio to main now that #601 merged #1062. It points at go-openaudio main and is harmless.
• The indexer cutover bounds (etlStartingBlockHeight/etlEndingBlockHeight, feat(api): add the indexer and flusher bounds the chain cutover needs #1028) and play routing (PlayRoutingHosts, feat(api): route stream resolution through named hosts during the chain migration #1029). The cutover still uses both. mustParseInt64Env stays because the ETL bounds use it.

Deploy notes

• The drop is irreversible and deletes any queue rows still in the table. Export them first if you want them for audit.
• Migrations run before the new pods roll out. An old pod with newChainQueueEnabled=true would log a warning on each relay until it is replaced. Its inserts were fire-and-forget, so relays still succeed.
• After the rollout, delete the seven env vars above from the deploy configs. Nothing reads them anymore.

Tests

• go build ./... passes.
• Fresh docker compose up -d --wait from the edited dump: go test -count=1 -cover ./... passes in 28 packages. In api, only TestFeedScheduledReleases fails. It fails the same way on origin/main without this change.
• Applied migrations 0232, 0242 and 0243 to a scratch database. The table is created and then dropped, and a second run of 0243 is a no-op.

🤖 Generated with Claude Code