@anlambert: can you please abandon this diff in differential?
AFAICT it has already been committed in swh-environment (if not, you can land it :))

(@olasd has already approved this, but it was still waiting for me)

Looks good.

This looks generally good to me.

great example! it has indeed all the needed features. Approved :-)

afaict, an easy fix for this is to just use "unset" at the end of the script on the relevant variables, instead of using "local" when declaring them

This won't work with zsh, and the pythonpath.sh script should be portable across at least bash and zsh, which are the shells commonly used within the team.
If you still have the problem, I fear you'll need to find a more portable solution than this :/

Gandi simple hosting unfortunately does not have apache's mod_proxy and, more generally, any way to do reverse proxying (short of doing that in wordpress, which is horrible).
So let's go for something like docs.softwareheritage.org, and serve the development documentation from there starting at /devel.

can someone tag this task with the most appropriate project?
ideally, no task should have no tag whatsoever…

*amazing*, thanks!

My current point of view is thus: we've been bitten by inconsistencies between primary and replica before, so I think the counts should run on the primary and get replicated through the standard means to replicas, even if that means stressing the primary a bit more.