Better monitoring of head.hackage failures
Currently the nightly pipeline of ghc/ghc> triggers a few head.hackage
jobs for validation of the HEAD compiler. However, we currently rely on human eyes to catch regressions in this job. Consequently, these regressions often go unnoticed for days or weeks. We could instead consider having some notification mechanism for these failures. Naturally, we would only want to trigger notifications on new regressions, which somewhat complicates matters.