We've had issues with articles being received from peers somehow ending up to be corrupt again. It's very intermittent and only happens for short periods of time (but during those short periods, hundreds of thousands of articles are received). It seems to be load related, but we haven't found the root cause of it.
SAB crashed when certain unicode characters was received in UTF-8 strings where it was not expected (notably in the =yend line. This is a SAB bug), and has been patched in SAB's decoders. That being said, it seems that this was not the only cause of problems.
We did implement fixes in our servers whereby we now validate the CRC of every single article containing yEnc (binary) data as they come in, which pretty much ensures that we won't ever be experiencing an issue similar to this in the future. SAB developers was kind enough to assist with tracking the issues down, and we use SAB's own verification libraries to do the validation too.
The fix went live on the morning of the 7th of April (last week Friday), and to date I haven't yet had a failed download personally. Nor are we aware of any clients experiencing issues with content newer than the the 7th.
Older content may still be subject to CRC issues. We only validate new content coming in, not existing content going out. We are however working on new server software which will work both ways.
If you can send us the NZB, we can manually remove the content at fault which should resolve the problems you're experiencing. We can't unfortunately manually validate a few billion files consisting of close to 1PB in storage, it will take weeks, if not months, to complete.
Sign up with Google
