Problems with Credits or validator malfunction?

Hi,

since nobody else did come up with the real reasons:
Validation has been a bit behind / irregular during the past days since
we've been busy defining "valid" and "matching results". Unfortunately,
although the validator is in place now, we're still hunting a bug that
may cause result being flagged "invalid" just because they have been
computed on a different platform than what has been chosen as the "canonical"
result. So if two Linux and one Windows machine return result files which
pass a first-stage validity check (we got a handful of files which contain
strange characters...), usually the two Linux results will match best and
consequently one of them is chosen as "canonical". The Windows result may
be too far distant from the canonical result, due to optimization/rounding
issues (still under investigation), so it will not get credit... Most of the
time, it's the other way 'round since Linux machines are outnumbered by
Windows boxes (at the moment at least, there are some clusters waiting to
start crunching :-)

As Bruce Allen said (in several other threads): we're still in an early
testing phase.
The next (Beta?) stage will see more reliable application binaries, perhaps a
wider range of supported platforms, and possibly another server.

Work is in progress everywhere, keep in touch.