soyuz is not fast-downtime friendly

[Tom Haddon]

Soyuz cannot currently cope with fastdowntime, which means a lot of workarounds to allow us to do fastdowntime:

- Disabe notifications of soyuz services in nagios
- Stop individual cron jobs to prevent them being long running
- Stop soyuz services before rollout
- Stop all soyuz crontabs before rollout
- Bring soyuz services back up after rollout
- Re-enable crontabs after rollout
- Re-enable nagios alerts after rollout

The idea of fastdowntime is that we shouldn't have to do any of these, which will make the process less error prone and quicker

[Stuart Bishop (stub)]

Last step of this is to remove archive-publisher from the fragile users list in database/schema/preflight.py

[Tom Haddon (mthaddon)]

process-death-row needed manual killing during today's rollout

[Tom Haddon (mthaddon)]

And buildd-retry-depwait on cesium.

[Stuart Bishop (stub)]

Bug #845397 is about making buildd-manager cope gracefully.

[Stuart Bishop (stub)]

Bug #815753 is about making scripts like process-death-row and buildd-retry-depwait connect as distinct database users, so they can be differentiated from other connections.