Fair enough — you're right that it isn't exploitable. The client sends signature_algorithms and the server picks from the intersection, so nothing modern ends up on SHA-224 unless it explicitly asks for it, which is exactly what my test did.
Where it still bites is compliance rather than security: organisations measured against the NCSC guidelines get it reported on every scan, and internet.nl states they expect to reclassify it from phase out to insufficient in a future update — at which point it stops being a recommendation and becomes a failed subtest. That's the only thing that really changes the calculus, and whether it's worth pre-empting is your call.
Same algorithm turns up on the mail server for the same reason, which is in the other thread.
