[libreoffice-website] unscheduled hypervisor reboot

Bugzilla and Askbot's databases have a handful of corrupted TOAST
tables, causing queries involving particular column & rows (such as
fulltext search, attachments, and search preferences for a few dozen
rows) to fail.

Bugzilla should be clean now; corrupted entries has been restored from
backups. Due to the granularity of our backups we're one attachment
short, and I asked the author to please re-upload. AFAICT no data was
lost (as far as Bugzilla is concerned) beside that single attachment.

Will proceed similarly for Askbot tomorrow. Assuming the ratio of
corrupted entries is the same, the service should be fully functional by
tomorrow European night.

Askbot's database is now clean as well. Issues in the ‘activity’ table
were causing hiccups in instant & scheduled notifications, as well as
some other views involving the activity log.

AFAICT the corruption is now fully resolved, and no data was lost.