I don't know whether just using an LLM is a problem. But in your case I would say the fact you used one and didn't indicate you did. If you indicated the answer came from an LLM, then the trust in the answer could be weighted accordingly by each user.
That's my opinion at any rate.
This is what has made me think they've likely given an ultimatum to the moderators to re-enable their subs "or else", because they're coming back thick and fast right now.
This will (if they get it right) reduce the number of moderators they will need to replace to get control of the remaining large subs. It would have been better if more mods took a stand. A lot of the larger subs are still private or restricted. So, we'll see what happens I guess.
The problem with replacing a large number of mods is twofold. Finding enough competent (although maybe that's not such a requirement) people to step into their shoes is just one. Another is that there's no real way to hide this. Those people that might not have had a horse in the API race might have one in the "removing my sub's favourite mod" race.
You could plug in a USB SSD or HDD and make sure the DB and other regularly written data goes there. That would pretty much remove the problem.
I would wonder how well it would perform. The limited memory and cpu power surely would make database access not great under even moderate load.