I have come to believe this is not a bug.
It may be that environment.d is very nice for systemd integration, but
it is not very granular. Any environment entered via that route will end
up in systemd user environments possibly even when it was not intended.
Scripts in Xsession.d on the other hand may well write over environment
variables but they are written in generic shell scripts and so can be
tailored to only set environment variables in specific circumstances. I
would guess (and I am guessing here) that environment.d gets processed
before Xsession.d, and so scripts in Xsession.d could choose not
override environment variables that have already been set if appropriate.
