wimlib

Have you tested the command on files that are inaccessible due to wrong reparse point information? Mabye that's the problem, although I hoped that wouldn't be an issue for deleting reparse points.

Thanks for your hint for only skipping Data Deduplication reparse points, that sounds a lot better than always just returning 0.

Thanks to your good instructions compiling was a lot easier than I thought from the first look

I did a short test which looked good and now I am applying a larger wim.

I am not sure if I completely get why you are not using that fix in general, but maybe this could help in some cases:
Only skip writing the reparsepoint if you are applying on windows AND the reparse point makes the data unreadable. And you don't mind it beeing kind of ugly. At least imagex discovered the error when I used verify/check, propably because rereading the files failed. Which also indicates that wimlib is not rereading the files for verifying the checksums.

Using NTFS-3G on a deduplicated volume seems to be a dangerous idea to me.

Yes, I did.

wimlib always verifies the checksum of data it reads from a WIM file, but there is no option to read back, for verification, files it has extracted. I think it's not worthwhile, and as soon as you assume that the filesystem can't be trusted to store things correctly, you have all sorts of metadata you would need to verify too. On Windows there are also cases where it's expected that not all files will be "readable" after extraction.