Sorry, I mean specifically when users try to stop the backup process. Not really network interruptions - those are automatically retried and continued failure would cause the backup job to gracefully end in a error state.
I wasn’t around during the early days (I’ve only been using Duplicati a few years) but I THINK it was when the devs added concurrency that caused the “stop now” function to cause database inconsistency in SOME situations. (Which has been improved in Canary channel, as I mentioned.)