How to organize a shared cache for qsys systems?

I have written a system to generate command lines and scripts/templates for various EDA tools. This will let me share the same same design across different implementations. The designs typically reside in a git (or some other VC) repository.

For Altera IP's only the *.ip is kept in the git repository. The cache tag is generated from a hash of the content of the *.ip and the Quartus version used as this uniquely defines a single *.ip file.

The build script will copy the '*.ip' file and run qsys-generate in the $cache/$tag location which will be added to the IP_FILE in the tcl build script. The next time the same Quartus version and same *.ip file is used it's referenced by the IP_FILE variable. This works pretty well.

Putting the *.qsys itself into a similar structure works too. But distributing the *.ip and *.qsys children of the parent.qsys in the same cache seem to be problematic as the parent.qsys logicalView XML tag seem to contain a partial relative sub-path.

Is there a way to make sure that only references to the child.ip or child.qsys will be present in the parent.qsys so that it can be located by the full absolute path specified in the IP_FILE?

Also how are *_hw.tcl files located and what type of references and relations between these and their respective *.ip and/or *.qsys files are stored in these files?

Or is the only solution to copy the full tree structure of the qsys into the cache and hash its full content?

BTW I'm using Linux only.