Replication breaks after snapshot/failover/restore cycle

The customer takes backups via GCP snapshot while the replica is shutdown.
The async cluster is then failed over promoting the host where the snapshot was done to master.
When the snapshot is then restored to the former master host, the replication is breaking after startup.

Replication error
Last_SQL_Error: Could not execute Update_rows_v1 event on table ecobee.SecurityTokenNoIntID202012; Can't find record in 'SecurityTokenNoIntID202012', Error_code: 1032; handler error HA_ERR_KEY_NOT_FOUND; the event's master log binlog.000062, end_log_pos 26775880

We have done several tests in attempt to fix this issue as seen in the referenced support ticket (#210916) with no luck so far.