The database is back, t-offline was kind enough to send a technician fixing the DSL-line. Seems that the DSL-Splitter died. I still think about a colocation in a datacanter for database-machine.
The database is currently readonly. I took the offline-time to do some database-maintenance and found a bad block in a 17GB-sized table.
I'm working to identify the affected rows and will then decide wehter to restore the old backup or restore all but the affected rows.