your error message will have to be a bit more descriptive: when user is useing nwchem, and wants to try out a new version, it will trigger this error. i'd add something like Please temporarily rename the self.home_nwchemrc during nwchem installation. It can be restored later for nwchem use.
Btw, we should cleanup the self.home_nwchemrc after installation (and to be on the safe side, also their i'd check if samefile(self.home_nwchemrc, self.local_nwchemrc)), so a user can simple rename the original home_nwchemrc

Very good suggestions, I'll take care of it.

While testing, I realized that cleaning up the ~/.nwchemrc symlink is not an option, since it may cause the test cases of other NWChem builds running at the same time to fail...

I've added a post message instead that reminds the user to clean up the ~/.nwchemrc symlink once all NWChem builds are complete. This will be a warning which is shown at the very of the output.

wow... this explains why I have been with the impression of so many installers having issues run in parallel; but then work fine if rerun again in isolation! I had assumed some common file collision, but an rm -rf on a dir like /tmp/NCBI* etc could also explain such cases...

if you agree, it would be nice to have the option to de-parallelize builds of the same package, because they are prone to bring up failure modes that the developers did not consider in their installation mechanisms...

How would you deparallellize (open an issue for that btw, it will be forgotten if we keep the discussion ongoing in here).

Lock files are not a good option imho, because we deal with shared filesystems very often...