As captured in RHCLOUD-40690:

We are seeing an issue where any time the platform MQ connect pod is kicked/redeployed/promoted, we experience WAL growth despite the heartbeat settings in our Debezium connector

Its possible moving to using a table for heartbeats vs the logical emit message is a fix but thats unconfirmed yet by debezium team (long thread here)

Currently there is a heartbeat fix script that runs in a CronJob that makes an API call every hour just to keep the connector active. We should take some time to review using a table for heartbeats vs the logical emit message

As an alternative, the heartbeat fix script could potentially leverage a Kakfa message vs requiring Inventory API to fix the issue. Either solution may allow us to onboard connectors for service providers (Specifically HBI) long before the migration and ensure we dont hit WAL growth issues on their DB.