Just to re-understand the problem:
Say you have 2 node cluster
- Node-1 is where you run the said long-running insert.
- Node-2 is replicating this insert through galera replication.
Now when the insert/delete is completed on node-1 if you execute another insert on NODE-1 … DOES IT HANGS?
OR
Do you mean when the insert/delete is completed on node-1 if you execute another insert on NODE-2 … IT HANGS
Latter is not an issue. Node-2 is busy applying the existing replicated write-set. It can’t allow followup write-set to commit till the replicate write-set is committed.
Former looks to be a problem.
If you are hitting a former case then can please share the reproducible test-case.