Follow-up from #632 and #621 It should be easier for us to switch/failover automatically if we lose a data centre. We should define what 'easier means, what kind of reliability we want to achieve, failover times, how we deal with desync issues, etc.
