Hooray! I think I figured it out!
The culprit did end up being CR9/CR13 side-by-side-ness, but not in the way I was expected: Since our (very old) Report Module was original intended only to support CR9, it issued a call to crpe32.dll's PEOpenEngine() on initialisation and simply left the Print Engine open throughout its life. This apparently causes subsequent calls to CR13 to hang in an installation environment (but seemingly not in a dev. environment).
Our CR13 support is kinda a graft onto the CR9 framework for now, plus it had been working perfectly fine in a dev. environment, so the open-once-and-forget-about-it paradigm seemed benign. I'm admittedly a little surprised that the behaviour in dev. is different from the behaviour in an installation, but I guess that's not outside the realms of belief. (In hindsight, I'm actually surprised that this ever worked at all in dev.)
Anyway, for posterity, I suppose the moral of the story is: If Open() hangs and you're reasonably confident it's not a permissions issue or a printer issue, check if the Print Engine is getting opened inappropriately, especially by a different version of Crystal.
Thanks again for all the help everyone!
Kevin