Quercus really is great stuff; it allows nitwits like me to develop crappy PHP-applications and to deploy them on Google’s App Engine. But when you combine the limitations of Quercus’ PHP implementation with those of GAE, you’re going to have to code around some problems you wouldn’t be facing when developing a “normal” PHP webapp.
One example based on my limited experience (while writing a scanner service to detect “foreign” objects in websites for my future wp-privacy plugin): I had a CSV-file that had to be downloaded & parsed. Normally you would fopen the remote file and than use fgetcsv retrieve all data line per line. Or, if you’d prefer, you could fetch the file with mighty CURL and parse it using str_getcsv. But those approaches don’t work when in Quercus on GAE; fopening remote files doesn’t work (blame GAE) and while you can Curl the CSV into a variable, there’s no str_getcsv in Quercus (yet).
So I did what any self-respecting non-developer would do; I cried for help on StackOverflow. Some of the advice I got there involved obscure tricks like using data-uri’s, fopening php://memory or using SplTempFileObject, but none of those solutions produced anything but errors. So no built-in CSV-parsing for me, but (simple) “manual” parsing of the CSV in a string. Not a huge problem by any measure, but I’m sure there’s a whole lot more limitations, if only for all those functions that rely on file handles. But at least we’re having fun, no? ;-)