"Bjoern Metzdorf" <bm(at)turtle-entertainment(dot)de> writes:
> Is this due to a corrupted record? If yes, how can I safely delete that one?
Hard to tell, though it seems like a good bet. A stack backtrace from
the core file produced by one of the crashed backends would tell us
more.
If you don't see any core file in the $PGDATA/base/yourdb/ directory,
restart the postmaster with "ulimit -c unlimited" environment to allow
core dumps to occur.
regards, tom lane