Just a short follow-up:
On site3, which is running since 4 months now, the same issue started out of the blue.
We changed the retention period for events and performance-data down to 30 days to keep the database small. This change was made about a month ago, the error-messages (internal tasks every 6 hours, see above..) started to show up about a week ago. So, here we have the perfect testcase for our issue.
This time I only changed the transaction-log upper limit to 20GB and waited for the vcenter cleanup-jobs to do its magic.
Success! The database is now 256MB smaller, the transaction log grew to 3.3GB.
Conclusion: for expiring data from the vcenter-database, you will need ~13 times its size in the transaction log.
I will now "shrink" the database as usual. Everything back to normal.
Kind Regards
Marko Engelmann