ACK 1dcbf0a

Agreed that it is better not to pin it, and that upgrading to a compatible version of the library is better, however, the issue here is that the CI ought to tell us reliably if the code we are changing still passes CI, and so any unrelated breakage adds to the overhead. So it might be ok to run with a fixed version on PRs and use stable in cron-based runs on master so we learn if the dependencies under us are shifting, without impacting the critical PR path (submit -> test -> fix -> test -> ...). How does that sound?

Yes, in theory I agree. But this is the only incident of a breaking regression in the Rust stable version in CLN I am aware of. Correct me if I am wrong.

Do you want to run against a pinned Rust version (which someone needs to update regularly) for PRs and pushed commits and introduce a new workflow with Rust stable (which someone needs to maintain) just because of this single incident? I feel like the introduced overhead by that added complexity in the CI is not justified. Especially because we are talking about a regression in the Rust compiler here.

Nah. Let's just keep it in mind for if it becomes an issue, you're right 👍