what you’re trying to say with arbitrary numbers pulled out of the air about the amount of “coders” and potential users

No matter how wonderful and loved a project (aka business) is, it has expenses. A project like Codeberg's expenses scale with the number of users and how they use the system.

This policy allows them to get rid of the worst offenders and point to the policy as their reason.

Agreed. However, at the scale they are operating at, I wouldn't be surprised if they end up using a LLM assisted AI implementation to give them short lists of projects to review as potential "worst offenders".