It depends on your use case - either you accept there is a write hole that the update wont appear until that host comes back online and other hosts can relay the change, or you add application-level checking to not return to the user in the first instance until the change hits at least another host.
In your scenario, yes. The two changes would be a conflict, but hopefully (if the user makes the same change) your conflict-resolution code would realise it is the same and merge them correctly. I fear I don't know how MySQL::Replication handles this, but it has to be application-level logic, so I assume it will allow you to override a default of, say, last-modification-wins.
In your scenario, yes. The two changes would be a conflict, but hopefully (if the user makes the same change) your conflict-resolution code would realise it is the same and merge them correctly. I fear I don't know how MySQL::Replication handles this, but it has to be application-level logic, so I assume it will allow you to override a default of, say, last-modification-wins.