I honestly think that is impossible/inadvisable, simply because of the potentially significant differences in scope/ complexity/granularity - essentially it is ‘false integrity’ you are trying to enforce uniformity when actually none potentially exists. You will end up with even more data quality issues than you started with.
I don’t think you could even be clear what is the ‘primary’ problem e.g is the patient’s ‘primary problem’ Wet AMD, leading to significantly impaired vision or vice versa?
My limit would be to try to make use of the problem-diagnosis archetype for each contextual problem list, to make it easy to copy between them (plus linkages) but accept that each discipline is coming at the issue from very different directions and that any ‘simple’ answer is likely to mess things up for others.