Risk Analysis Should Elevate Risk if Most Tests Did Not Run

i.e. install failures.

May require some very explicit install handling?

Possible simple solution: Lookup a successful job run, see how many tests it ran, make sure we're in range of that.