If a circulation manager has a large number of configured collections (most common when there are a lot of Overdrive Advantage collections) then app server startup takes a long time. This can be annoying on its own, and it can also interfere with other ideas like https://jira.nypl.org/browse/SIMPLY-967, which save memory at the expense of creating app server processes more often.

It should be possible to parallelize the creation of API client objects on startup, similar to how we parallelize the process of asking all the APIs about loans and holds for the active patron when generating their loans feed.