DistributionSourcePackage:+filebug timeout: bug notifications are calculated in-request

https://lp-oops.canonical.com/oops.py/?oopsid=1670EA4372

https://bugs.edge.launchpad.net/ubuntu/+source/linux/+filebug/f8394cfa-9a88-11df-b34a-002481e7f48a

SQL time: 9336 ms
Non-sql time: 5466 ms
Total time: 14802 ms
Statement Count: 541

The non-sql time is a bit misleading due to the bug with oops' and sql statements that get cancelled sometimes not accruing as sql time.

Got through to generating mails-to-send but timed out at that point.

There are 6 inserts per recipient, and 0.4 seconds per recipient.

If I was hacking on this, I'd remove the flush(), migrate any dependent doctests to unittests that can be more tolerant of ordering changes - or order the returned data in the test.

Ideally we'd do this work out of transaction - you could use the jobs system to do that today, but the flush is probably a good enough band aid.