[Glass] GS 3.1.0.5: Fatal Internal Error, 5015, 15621, IO Error (RepRead Error)

[pieter at nagel.co.za]

Just an update:

So far it seems as if the bug goes away when we roll out the
STN_TRAN_LOG_DEBUG_LEVEL setting - hard to say for sure though with a
Heisenbug.

As part of setting STN_TRAN_LOG_DEBUG_LEVEL, we also have to switch
tranlogs from /dev/null to real directories.

So I strongly suspect that tranlogging to /dev/null is the root cause of
these RepRead errors.

I won't be surprised if the code paths around that are not as well tested
- I think we here are a bit unusual in the GemStone world in spending so
much time in that "mode" which other people only use occasionally to speed
up bulk imports.