The RUC2 crashed late in the fcst runs last Fri at 12z, 15z, and 18z. The
problem was a very low water vapor mixing ratio value at the top level in
the model, which
eventually became negative. A piece of code designed to prevent this in the
model had left the top level unprotected, which is where the problem occurred.
The actual crash was in long-wave radiation, which did not appreciate being
given a negative water vapor path length.

At 21Z on Friday, the RUC2 was forced to use the 9h Eta as a first guess,
which temporarily got rid of the dry spot at the top level, and the RUC2 ran
reliably for the weekend. This morning at 14Z, a fix was put in to extend
the check against such problems to the top level so that this problem will
not occur again.

Stan