I’d be in favor of it. With a federated platform it’s impossible to keep someone from spinning up per-interaction accounts so we essentially need to deal with the costs of allowing anonymity anyway. It could be implemented in such a way that your host instance can connect the interaction with an account if there was some advantage to that but banning an account created to post something offensive and a moderator removing an anonymous post that’s offensive are pretty much the same action - and if a particular account was egregiously abusive there are technical ways we could allow the host instance to ban the user without compromising anonymity for other unbanned anonymous posts.

Oh, anonymous DMs should be absolutely disallowed. That’d just be an open invitation to dick pic spamming but, as nearly any open woman on the internet will tell you dick pic spamming is a constant issue.

Basically, it really wouldn’t make any kind of abuse easier and even if it was significantly abused to bypass needing to do account sign-ups for trolling there are pretty simple technical solutions to fix it.