Hard to remember them all now, but definitely had a number of bugs I've ran into, most able to be worked around. For example a recent one was a huge slowdown in a simple query that should be optimized[0], still have no great solution for it or attached to any roadmap.
One big one is at some point you need caching on a query level, and for that you have to purchase enterprise. I suppose they need to make money somehow, and it's a smart level to do it because it's almost impossible to not need caching at some fairly early stage. It would be nice if they let you test it out without "Contact Sales", and while there may be some way to wrap it from above, I don't think it's very easy.
A meta critique would be the project seems to have less momentum than Prisma. Releases aren't as frequent or large, and I'd have much preferred they double down on PG and keep adding query abilities, performance, etc, instead of expanding database engines.
I find that with new tech, I'm more interested in what it can't do rather than what it can.