the accumulation of hash keys in the "boolmap" structure is going to have the same effect.

Indeed. But once I've found a set ready for removal from the queue, I have the resource identifiers (formerly boolean indexes), and so I know where in the boolmap to look, but I still need to locate the resource identifiers (formerly set indices), in order to remove them. So now the values within the boolmap have to also become hashes rather than arrays to avoid those linear searches.

When adding a new "set" to the queue, what if all the booleans in this set are already true?

The scope for this is limited, but still possible. The solution is that it is relatively easy to check whether there are any outstanding resources for a given set. If not, it can be processed straight away and doesn't need to be added to the queue.