On a personal note, I very much agree, and we covered this in the initial actor's pitch in the Actor reentrancy section. swift-evolution/0306-actors.md at main · apple/swift-evolution · GitHub and in the future directions in the same document: swift-evolution/0306-actors.md at main · apple/swift-evolution · GitHub
It is somewhat unfortunate that our actor runtime does not provide such flexibility (yet?), as it is something either subconsciously expect or actively need and can't really achieve without reaching to streams and manually reimplementing queueing...
I believe the solution would be to allow users to opt-into non-reentrant behavior for methods which are in need of such crucial guarantee, but otherwise reentrancy could be permitted - as was pitched in the evolution writeup above. Have a look and say what you think. Deadlocks are a concern, but in my opinion way easier to diagnose and debug than that one once-in-a-blue-moon weird ordering that causes a concurrency bug in what otherwise is a nicely synchronized actor.
Having that said, I'm not really speaking on behalf of the concurrency or core teams, but based on my experience on working on actor runtimes as well as swift's take on it. The actual solution may need more research and we'd be open to hearing alternate takes I think.