So the answer to this question is normally pretty straight forward, being " collimate without a diagonal". The basis being that there are many different Diagonals and some are not ' straight ', nor even adjustable. Manufacturer collimation instructions sometimes specify that collimation is to be done without a diagonal. So we dutifully collimate our SCT, CC, etc to perfection, then stick on a diagonal and the resulting visual star test is a mess. What to do?
Do you tweak your diagonal screws, assuming it has them? Or do you re-tweak the scope collimation to give the best possible star test, effectively compensating for the diagonal using the scope settings? Of course in that case, if you move to imaging you need to re-collimate again.
What factors inherent to Diagonals and Focusers can alter collimation:
1. Body and/or barrel misalignment in manufacturing, thus can result in tilt at the EP end.
2. Some focusers do not run straight (the GSO RC focuser being a good example), often the angled racking path is not due to poor focuser tube running alignment, but the way the focuser is attached to the back of the scope.
Those are just a couple of examples.
What are your views, compromises, techniques, etc that you use to optimise collimation?