Here's the results of my study in thread pooling in Javascript: https://gist.github.com/halfak/5e0c4951c52f57ff3320aff9b51b757e

Essentially, a system like this allows a developer to work with a thread pool without minding the number of threads. It's a simple asynchronous pattern that allows for filing jobs and getting a Promise() in response. The system also uses Deferred's to manage it's worker's activities.

Based on this, I needed to implement something slightly more complicated in ORES, I needed to both manage a set of workers *and* batch requests as they come in. The result is captured in this edit: https://meta.wikimedia.org/w/index.php?title=User:EpochFail/ArticleQuality-system.js&diff=18481658&oldid=18414783

Essentially, it implements a similar worker pool as is described in the threading example above. But it also draws tasks from the queue in batches. Within a batch job, individual Deferred objects are resolved. This allows a user to just make requests as though they were going to ORES one-at-a-time. The interface for ArticleQuality.oresScore() has not changed at all. The scoring jobs are then queued, batched, and distributed transparently so that the end-user developer doesn't need to keep track of anything but the Promise.

When testing this out on pages like https://eu.wikipedia.org/wiki/Wikipedia:Wikipedia_guztiek_izan_beharreko_artikuluen_zerrenda/3._maila (lots scores needed to annotate the links on the page), the performance is plenty fast.

(Could you add a project tag to this task? Thanks in advance!)