Improve Concurrency Model

[dechristopher]

The current work queue concurrency model works fine. The only optimization I can see is if there's some sort of global work queue for all zoom levels with a consistent set of n workers that aren't re-spawned at every zoom level.