The error is rather subtle it has these effects:

• New replicas would not be able to connect with a gtid in the past.
• On a maxscale restart, replicas already in sync, would appear not to connect until the next update from the primary, but would otherwise function correctly.
• A maxscale crash leaving an incomplete transaction in the binlog file being written, would allow the writer to append new transactions on top of that trx, breaking replication.