When do I know it's time to ship and stop polishing?

Define sticking as a useful repeat action over a set period, rather than another login. Bring in one more small batch from the same Reddit audience with the product unchanged. If they complete that action too, you have a stronger reason to spend time on distribution.

10% of a few hundred people sticking around without you pushing them is a decent signal. I'd talk to that 10% before building anything else. Ask what they use it for and what they'd miss if it went away. If the answers are consistent, you know what to sell, and distribution is the next problem. If every answer is different, the product probably still needs work. I ran a SaaS for about 7 months and spent too much of that time adding features nobody had asked for. If I did it again I'd stop building whenever no current user was asking for something

sometimes shipping early and learning from real users is better than spebding too much time polishingevery details

I’d pick one job the returning users come back for and fix anything that blocks it. Then bring in a small new group and watch whether they reach that same moment without help. That gives you a clearer next step than adding another feature.

One angle the others did not cover: before choosing between polish and distribution, look at the 90 percent who did not stick and find the exact step where they stopped. Polish is worth it only there. In my app the numbers once showed that almost half of the people who made a song never pressed play on it. The fix was tiny, the finished song now starts on its own, and no amount of polishing other screens would have found it. If the drop is spread evenly, the product is fine and it is time to distribute. If it is all at one step, fix that step first.