Issues with Claude Sonnet Variability in AI Content Generation

Hi all - curious if anyone else is deep into using Claude Sonnet 3.5 / 3.7 with the AI web research and content creation / manipulation tools. Over the past 2 weeks or so I've noticed that when I "re-run" prompts, I'm not getting materially different results like I used to (primarily using the content creation / manipulation function for email copy). It seems like each re-run only changes a handful of words but generally keeps the same messaging. I haven't really changed the prompt over this same period of time, so I'm wondering if there's something else going on that I'm missing... maybe need to start a fresh table. As I'm sure many of you have experienced, these prompts don't always get it right the first time, so I have relied on more variability between each "roll." Let me know if anyone else is experiencing the same / found a solution! I know there was an issue with re-running prompts last week, which seems to have been resolved... Perhaps these issues are connected in some way?