0 rows affected on delete in galera cluster that actually deletes data

I know that writes on multiple galera nodes should be avoided or handled with care, but I've seen that simultanious deletes can result in "0 rows affected" even tough the delete is executed on the cluster and probably as a result of the result, the query is not written to the binary logs (setup to row based replication).

It seems wrong that a result of a query can be "0 rows affected" and nothing written to binlogs when it actually deletes data - it should result in an error somehow so the application can handle it.

I've created a simple docker setup with 3 nodes, a bash script to start it up and a bash script to run the query to make it fail...

https://github.com/anderskvist/galera-debug