> I've thought about this, but I don't want to do this, and not only
> because it's work.  It also ties us more closely to Github, which I've
> always strived not to do, partly because free-software folks have
> reasonable objections to Github, but also because I've learned the hard
> way that tying oneself to someone else's platform is a long-term risk.

That's definitely true, but in case of hosting change (to GitLab or
elsewhere) one would simply rig up a new batch job to export from that
service instead. The issues and PRs are not an official channel, so
people scavenging things from them could reasonably expect to encounter
some inconvenience :)

> Almost all of the work on the pull requests themselves is captured in
> the official Git repos and on the mailing lists, partly because I
> deliberately don't make substantial comments through the Github UI,
> routing them to the mailing lists instead.
>
> The thing I worry about more is people advertising their own Git repos
> for intermediate work.  It's reasonable to do, but it makes it harder
> for people reading the record to follow the context.

The simplest fix is that you as the editor don't accept substantial PRs
without announcing a new draft to go with it. This should work well in
practice since you are the only editor. Enforcing group standards would
be harder if there were multiple people as gatekeepers (or if we were
operating by consensus).