The data directories used by SPLIT_DATA on a shared NFS volume looks like these:

ls -hlan
total 12K
drwxrwxrwx. 4      65534 65534 4.0K Jan 20 21:30 .
drwxrwxr-x. 7        185     0  103 Jan 20 21:08 ..
drwxr-xr-x. 3 1011640000 65534 4.0K Jan 20 21:07 xa-load-2
drwxr-xr-x. 3 1011640000 65534 4.0K Jan 20 21:13 xa-load-3

As the migration pod needs write access to the directories, it will only work when having the same uid.

See https://docs.openshift.com/container-platform/3.11/install_config/storage_examples/shared_storage.html , it would probably be better to use the supplemental group Ids (and make the directories owned by such specific group, with having them group-writable)