Overnight, the ImportantSlaveNotRunning alert went off.  The alert
happened because the jadeite-release builder (a canary slave) had
missed two scheduled runs since the regular 11:00 AM run on 5/29.

The alert was correct that the builder hadn't run, but the message
wasn't relevant:  The slave had been turned off in goldeneye and wasn't
meant to run.  We should figure out how to make this alert not go off
in such a circumstance.

Here's the relevant text of the most recent message:

Alert Details
------------------
Description:
An important slave has not run in 1 day.

name: ImportantSlaveNotRunning
current value: 0
threshold: Lt(1) for 3h
alert fields: {, metric:slave_config=jadeite-release,metric:master_config=master-release}

sent at: 2018-05-30 05:49:28
active since: 2018-05-30 02:48:58 (3 hours 0 mins)