Mea culpa, I wasn't on the most recent revision.

I can't reproduce this (I actually checked it before merging). Did you recompile
on the up-to-date default version?

--random-seed still appears in the --help output. I suggest grepping for it.

Merged and pushed.

Not sure if I wanted, but I did (of course after running the test locally and
fixing everything :-) ).

I left a few comments on bitbucket. The code looks good to me. Do you want to 
enable pipelines for this repo to see if all tests pass and all standard configs 
still run?

I prepared a pull-request here:
https://bitbucket.org/SilvanS/fd-dev/pull-requests/24/issue649/diff

No experiments have been setup yet, because I'm not sure if that is necessary,
after a code review. After all, the "global" RNG object now only lives somewhere
else than before. However, they could be a good safety-check (but would requires
some work to setup all the configurations...).

This is a follow up issue of issue648.

All current users of g_rng() should instead use the helper functions in
rng_options to add their own random_seed option and based on it get their
own/the global RNG object.