Hi,

thanks for the patch.

I'm not thrilled about that approach (arguably Guix already does it wrong
anyway).

But since the manual of pam_limits does describe that one should use it
like that, I have applied it as a stop-gap fix to guix master as
commit 0bf7d34d77ffca40be9e04586195054e9f2c7a13.

Long term, we should really make pam entries first class and show up in the
operating-system record--that's what they are FOR: to let the administrator
(and thus the organization) choose how they want to do user
authorization/session handling etc.  Why do we decide for them?

Bug report kept open for obvious reasons.