I would urge you to slow down and test more thoroughly, and do more comprehensive testing before posting more about it. Intermediate results that something (smallish) worked are not useful because they don't prove anything.
You could start by using the same content that's trouble, with cp, and then you could do multiple cp in parallel to subdirs. Write a script (don't just type), and write comments about why you are making choices. This includes where the fs was created, options, and why those are valid choices. After a test run, unmount, mount on linux and fsck there. If you can get it to fail, move to having the script generate synthetic data, and get it to fail, and then you can post the script. Nobody can replicate anything you are doing now. It seems that you are testing on the same disk that's been trouble. Almost certainly, you should also be testing on a different disk, with a different computer, and keep rigorous notes about what you did, where the filesystem was created, essentially writing a blog post about your testing. It would be good to figure out how to post that someplace and do that -- the details in multiple messages are too much for the list and too hard for people to follow (e.g., I'm not really trying).
