The worst kind of the (Oracle) bug…
is not ORA-600 or ORA-7445 or anything that starts with ORA-nnnn at all (not even data corruption!). In my book the worst bug you can hit on RDBMS server (of any kind, not just Oracle!) is the one, that will cause RDBMS server to return the wrong result set – without raising any error at all! These “little greens” are extremely hard to trace/detect in the first place; it usually takes someone who knows data intimately to notice the silly results (and DBA’s nowadays are usually not that familiar with actual semantics of data they manage).
One such bug we recently hit is Bug#4604970 – Wrong results with ‘hash group by’ aggregation enabled. The bug was introduced in 10.2.0.1 and it was fixed in 10.2.0.3.
Unfortunately, we’re stuck with 10.2.0.2 on production server, until we’re ready to patch our 10.2.0.2 production database to the release 10.2.0.3, we’re more or less forced to use the only acceptable workaround suggested in above Metalink note and that is, to turn hash-group-by feature off, by setting “_gby_hash_aggregation_enabled” to FALSE.
Pointer to another bug related to GBH…
I would like to point to another bug related to group by hash. Not as dangerous as this one, because it’s not silent, it fails with ORA-600.