Research question — agent memory + reliability in production

[abhishekkmca]

Hi — I've been looking at CrewAI Studio and it's a really solid project. I'm doing research on the infrastructure gaps developers hit when building seriously on top of CrewAI — things like persistent memory across restarts, task reliability, audit trails.

I imagine building Studio gave you a clear view of what CrewAI doesn't provide out of the box. Would you have 15 minutes to share what you had to work around? I'm trying to understand the real pain before building anything. No pitch — just research.