What kind of error did you get?
I think the reason for the low CPU usage in the first pass is that we use single-thread Avisynth. A multi-thread version exists, but I'm not sure by how much the encode will speed up. Plus we'd have to modify the script.
By the way, as long as you kept the files .stats and mbtree.stats from your last encode, you wouldn't have to redo the first pass again, just the second.