It is imho a system stretched to limit with database operations. Restoring backups in such a crowdy setup always is at risk to fail. However the rebuild on reboot option did fix it.
I don't think this very special usage pattern gives much hints for standard users ;.)
Come on, Boris. With all due respect.
Having two stations with 18 sensors each ([Temp/Hum in, Temp/Hum out, solar, UV, wind, rain,] 5x soil moisture, 3 x extra Temp/Hum, Lightning, 2x PM2.5) is not too unusual (given that MB supports now up to five (!) stations).
Some simple mathematics:
a "classical" station has 7 sensors (WH2320E or HP2551 w/ WH65). 7 x 5 = 35, and I have 36 for 2 stations. So it's inside the warranted range, isn't it so ? And as for the http requests, there are 5 weather networks and 2 websites ...
The top model of the MB family, the MB Pro, should be able to manage that. And I have already spread the http load over two MB Pro machines.
By the way, I have already observed occasionally that strange behaviour earlier with only one station.
After some time the system then purges the RAM DB and starts from scratch again.
As I'm saving (exporting) my minute based data hourly, I'm not losing much when that purge occurs.
So I'm not sure it is only due to "extreme" loads on the database that this database corruption happens.
Right now it's working again and thanks for the swift support.
Keeping my fingers crossed that there was/is no deeper issue behind.
At least I know now how to handle such a situation in the future - and am happy that I invested in a backup machine (which interestingly didn't go down even though it mirrors the database i.e. also runs on two stations and even took over all the additional load from the one with the DB error for the repair period).