I'm not sure what our use case is. I discovered this when I was looking at our project's util library, and we have a Queue subclass that overrides _full to handle the scenario where the queue shrinks. I'm guessing it's being used to dynamically adjust the queue size for performance reasons. Consider where you have a queue of items, but depending on the run time environment, that queue might need to be more or less limiting. Maybe the queue is larger during the day when requests are heavier and shrinks at night (so it's not accepting more items than it can handle when day breaks).

Surely that's a contrived example, and I suspect our actual usage is more concrete. I'll see if I can learn where we're using that functionality.

If shrinking maxsize is not allowed, that should be enforced in the interface (such as with a read-only/increase-only property) and documented. If you would rather proceed that way, I'll draft the patch for that instead.