We are getting H12's from Heroku when Azure takes >30s to respond to a credential request. Let's manually time that out at less than 30s and return a suitable (5xx) error message to the client.

Let's put, say, a 6h expiration on that cache, too.

Commits pushed to master at https://github.com/taskcluster/taskcluster-auth https://github.com/taskcluster/taskcluster-auth/commit/db83723f26a49103103679b87b0f6b552d36a300 Bug 1434369 - don't try to create tables / containers every time This operation occasionally times out, so let's try to do it less often. https://github.com/taskcluster/taskcluster-auth/commit/e438fc490d2b187e40f7673be9b5c59dac53a998 Merge pull request #130 from djmitche/bug1434369 Bug 1434369 - don't try to create tables / containers every time