If Descript fails to automatically re-sync the audio and text, it will distribute the words evenly over the audio span. One way to work around this is by making changes to the script that trigger the re-sync process.
If you want to re-sync a large string of a text - a sentence for example, you'll have to work quickly to extend the re-sync range by changing the spelling of many words... kind of like this:
We are always working to improve alignment accuracy, but hopefully these workarounds can get you moving in the right direction.