I strongly believe that for stable tasks not only models should be monitored in production, but there has to be a business process which takes any detected mist

I strongly believe that for stable tasks not only models should be monitored in production, but there has to be a business process which takes any detected mistakes of the system and feeds them back as training data, creating a "human-in-the-loop" system even if the vast majority of the data is handled automatically - as these mistakes have a much stronger useful signal for updating the weights as any random sample.