Simply: don’t auto remove a filler word if there is audio on another speaker track at the same time. Or only replace with silence on that one track if you want to get fancy. A similar idea is one of the top requests for Descript Classic, but it is a real problem in the new storyline mode.

The use case is I have two podcast speakers in separate tracks. Sometimes we talk over each other and one person might say a filler word. Descript will currently snip and obliterate both tracks to remove the filler word, which messes with the flow of audio on the good track.