If a failover occurs the last known GTID of the old master server is ahead of the about to be promoted master, a warning should be logged to indicate that transactions might've been lost.

In practice this seems to be a problem when semi-sync replication turns itself off and the configuration that otherwise would provide lossless failovers ends up being a lossy failover. An additional improvement would be to warn whenever the semi-sync replication turns itself off (i.e. Rpl_semi_sync_master_status == OFF) to notify admins that if a failover does occur, it might not preserve all transactions.