Why do we have a split between Practitioner and PractitionerRole?

In tracker item #19698, Alexander Henket made the following argument around the use of Practitioner and PractitionerRole:

Practitioner is sort of useless because you cannot reach specialism nor organization from that resource. There seldom is a requirement for stating "some doctor with certain qualifications picked up meds" There usually is a requirement for stating: "cardiologist X from organization Y picked up meds".

The very reason for the existence of PractitionerRole is that it is the only way to know in what capacity someone works in an Organization. There is no useful link between Practitioner and Organization other than PractitionerRole. If it is important enough to make it known that a Practitioner picked up meds, then it must be relevant in what capacity this Practitioner was acting. For that you need access to PractitionerRole.

If you take Alexander's argument and abstract it, then I begin to question why we have a separate Practitioner from PractitionerRole. If everywhere I want to make it known that a Practitioner did something, then I must want to make it known in what capacity the Practitioner was acting. Therefore, why do we split them up? Just add the Role parts of PractitionerRole to Practitioner as a 0..* block and now we just have a Practitioner resource that conveys our needs.

This would also help in all of the mapping exercises I am doing where I have to choose between Practitioner or PractitionerRole depending on whether the source information has an equivalent of role and organization or not.