Some people (including me) would actually love to see hydra bumping nixos-unstable being replaced by a bors-like system.
There is basically no drawback (no commit will appear in unstable unless hydra passes anyway, the biggest drawback I can see would be that nixos-unstable-small would no longer exist). And it would mean that all PRs ready for merging would actually land even if another PR were to break hydra, thanks to bors’ dichotomy-based bad PR identification strategy. This incidentally means that the need for nixos-unstable-small would probably just go away anyway, as I expect such a hydra to move much faster as it’d basically have automated rollback.
Adding other checks would probably make sense later on, but as a first step it seems to me like this would be a net win, plus it’ll prepare the infrastructure for when we will actually want to implement more checks later.
The one big issue is actually infrastructure, as I seem to understand plugging into hydra is not actually easy, and we probably don’t want to be rebuilding all these things twice. But maybe building enough to check hydra would be green is not actually that expensive and it could reasonably be done outside hydra?