Chatting with @alessandrofelder we thought that running the benchmarks periodically could also serve (albeit indirectly) to detect dependency breaks.
However since this is a somewhat indirect way to test this, we suggested adding an additional CI workflow to regularly run the workflow tests and detect this kind of breaks.
Stretch goal: add accuracy checks (we want to flag if accuracy / performance changes when a dependency changes)
Chatting with @alessandrofelder we thought that running the benchmarks periodically could also serve (albeit indirectly) to detect dependency breaks.
However since this is a somewhat indirect way to test this, we suggested adding an additional CI workflow to regularly run the workflow tests and detect this kind of breaks.
Stretch goal: add accuracy checks (we want to flag if accuracy / performance changes when a dependency changes)