you obviously attempted to use imputil to implement non-standard import behaviour on top of the standard storage system -- while we've used it to implement standard import behaviour on top of non-standard storage systems.
No, I tried to make the imputil approach work as replacement for the standard builtin importer.
I'm confused. earlier, you said (or rather, I think you said) that you looked at imputil to see if it could "handle the problems you had at the time"... and now you say that you tried to use it as a drop-in replacement for the "standard path importer". I must be missing something here...
After I got that to work, I added some caching to avoid duplicated stats. The resulting importer was around twice as slow as the builtin one for the following imports:
# the default one Python does at startup, plus: from mx import HTMLTools,DateTime,ODBC
This is a pretty common setup for my scripts, so its preformance is relevant to me.
did you try stuffing all your PYC's into an archive file, and running them from there?