Drafts cases from the week's activity, then answers the tests, a few
suites at a time. startedAt is when the request began (default:
now): the runner's time limit counts from then, and tests it can't
reach in time carry on in fresh requests.
Brings attack cases up to date with newly confirmed answers, then
records each suite's test on its live model, so the hub shows them
running. Returns the tests' ids for run.
The weekly check, in the two halves a request can split it into.