|laziness, impatience, and hubris|
I would try to avoid mixing threads and Win32::OLE, as I fear to omit relevant parts of the dance described in the documentation for Win32::OLE->Initialize().
I'm aware that running multiple instances of OLE in separate threads may not make for complete isolation even using require, but the failure appeared to be rooted entirely in Perl code rather than the underlying DLL's (which by default are initialised for multithreading).
Hence requireing it into each thread does appear to provide isolation at the perl level; at least as far as the few trivial tests I've done go. It certainly fixes up the free to wrong pool error that was the cause of his headline problem.
Over the years I've found several modules that "aren't threadsafe" work just fine if you require them into the thread where they are used rather than letting them be duplicated there at thread creation time. Usually I advocate only requiring them into a single thread, at which point they do not see any difference between being one thread is a multi threaded process and the only thread in a single threaded process; but in this case I thought I try requiring into multiple threads and it seems to work. (Maybe I should go back and try the same thing with some others; like Tk).
I guess we'll find out if the OP ever gets back to us.
With the rise and rise of 'Social' network sites: 'Computers are making people easier to use everyday'
Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error.
"Science is about questioning the status quo. Questioning authority".
In the absence of evidence, opinion is indistinguishable from prejudice.