Now resolved it seems. Interesting failure though:

We have identified a critical cooling failure in our Frankfurt datacenter. We are currently in the process of shifting traffic to our backups sites.