First, it's not clear that the failure to fall back to losing forward security in the event of one-time prekey exhaustion is a decision - if you simply error in

First, it's not clear that the failure to fall back to losing forward security in the event of one-time prekey exhaustion is a decision - if you simply error in your client code when you try to get a prekey and there aren't any, that could also be because you didn't understand that libsignal will perform that fallback. But the more important point is that the spec is explicit about the importance of trying to prevent exhaustion of one-time prekeys, and their failure to do that is absolutely an issue. I was trying to talk about the consequences of that issue, rather than saying that the decision not to fallback is inherently itself a security concern.