In a dual master replication, you can easily have issues if you do it manually without a cluster or something else.

There is not much difference between repairing a master/slave and a master/master replication. However, you should take care of one evident thing: when you're working on an issue on Master 2 for example, you absolutely need to stop the slave replication on Master1!

The advantage of doing this is simple; if you make mistakes on Master2, they won't be replicated on Master1. This helps you to keep your Master2 node as close as Master1 if you can correctly repair it. However, if you fail, you can recreate from a dump of your Master2 without impacting Master1 data.

That's why, in a dual master replication, the first thing to do is to stop the slave replication! If you are not sure which one to stop, it's preferable to stop all slave replications instead of modifying unwanted data.