To the first I would say that if you rely on frame action to get stability, you normally would accept some degree of believable fixity to make the overall lateral resisting system more slim. What can be a believable degree of fixity there, there are many opinions, but a nonlinear analysis revealing the degree of the cracked zone can give an idea of how you need to model stiffness, or you can check in different articles etc what is said about. Some go to design based in an entirely elastic scheme (even if with some reduced stiffnesses) with almost unnoticeable effect in serviceability and maybe something more at limit strength. Since for use it is feasible, it is done (may or not meet some particular codes).
When you rely in an entirely separate lateral resisting system, say shearwalls or braced frames, I find more acceptable to model as hinges the supports at exterior columns, specially in not highrise buildings where the lateral displacement may be reduced with such systems and so any effects of the notional hinge small. In spite of that here in Spain has been common (for framed, not particularly flat plate, buildings) to reduce much the moment at exterior supports respect that produced by elastic fixity. Again the visual effect in common dwelling buildings is nil, and you see there more "crack" due to the separate concreting of column and floor than mechanical cracking from overstress.
Respect transfer beams with important loads, yes, it is common to make use of hinges or reduced fixity at ends to ease some constructive details. You may have also cases with elastic fixity and well, as I have practiced and seen them have not resulted as bad as one could at first sight think. Taking unto account continuity also helps the lateral resisting system, for here, the buildings not being very tall, definitions of lateral resisting systems as separate entities are not as differentiated as in the US, and everything works together. I also have designed hinges for ends of steel multispan transfer beams to deliver forces to concrete through an anchor plate backed by an array of studs in columns, and bearing in sockets in walls, and again, no problem. The big stiffness of the transfer beams has much to say on this since it forces not much the respective housings other than in the devised manner. It is in the same manner that a shearwall protects weaker beams with ends designed as hinges, by diminishing the displacement excursions.
Yor last question is maybe where we may produce a more tolerant design, since we have plenty of support. You may design the node as simply supported, you, from your questions, are aware of the cracking issues, so it is a matter of preference and, for dwellings at least, minimum geometrical rebar should control the issue to limited proportion. In any case since the trend is to mandate quite significant minimum rebar anywhere and specially where tension appears, I have sometimes (much times) used elastic fixity atop RC walls, and I remember once it was to the surprise of some insurance reviewing party that was criticizing bottom rebar of the end spans without thinking I have done such thing.
Maybe after this account of my experience, the bottom line is that you can take almost the whole gamut of fixities and still have satisfactory buildings; that some may sastisfy some particular code or preference is an altogether different thing.
