Is this expected that some systems work and others don't?
We've observed such behaviour during tests. For now it can be described as follows:
If the module has been built via dkms on 7.4 system with kernel version < 3.10.0-862, then dkms will fail to rebuild the module after kernel update, BUT weak-updates will kick in and the module might keep working. This is not supported as we don't know potential pitfalls of such approach.