quite often people use reds to "grab over each other", taking items from any of two belts (4 products on 4 lanes, or even up to 8 items like science packs from two belts with beltweaving) and/or dropping items from the assemblers back onto one or the other belt (eg getting items from 3 lanes and dropping products on the fourth lane).
edit: writing and re-reading this, i just got an idea how to use it on fantario, getting up to 6 items from 2 belts and dropping products and the surplus junk on two other lanes of the same two belts, provided I'll get longhanded burner filter inserters :-) LOL
the halved speed can be made up by using two of them in parallel, and long blues would of course be some kind of cheating, but all new additional inserters would be kind of "mild cheating" anyway :-)
no advantage of yellow over red ? there are some advantages of long yellows above reds: although they are slower, they don't need automation research, they need less power (same 400W drain, but only 13kW instead of 18kW in use, 30% less) and they are cheaper and easier to make (reds are made from a yellow and thus require an additional assembler and almost twice as much iron, 7 instead of 4). of course, those advantages are only important in the absolute beginning.
i see the same problem "the other way around" with long blues: why would i use reds when i can have faster long blues? :-) they need additional copper, but are similar to make like reds and then are faster.
this is also why i like to be able to select which inserters i use (schall, kry, etc) and independently which angle/range changer i use (bobs, cursed, etc). but too many thoughts about this will probably be not worth the effort since this would turn the planning of mods and mod combinations into the "real" game instead of playing and using them in some modpack :-) most important will probably be my (always different at the time) "current" needs, eg having additional long and filter burner inserters as well as some leeching capabilities when playing Fantario :-)
the exact drop location within a tile is most important for angled locations (0, 90, 180, 270 are no problem, and eg 22.5->67.5 or 22.5->112.5 are no problem either, but 0->22.5, 0->45 or 22.5->45 causes slow extensions and retractions), thus for side inserters this only matters in special cases (when dropping on corners of belts). i stopped using side inserters many versions ago when i wanted to unload from a belt to a chest with three "parallel" inserters (maybe even stack inserters, one straight and two with drop slanted by 45 degrees), not wanting to use a normal inserter and two reds which would have made the setup slower and also larger from 1x3 to 2x5.
speaking of different properties and combinations, etc: another new mod might be interesting where inserters are built from parts, having a base (burner, electric, heat, whatever), an arm (short, long, extralong, extensible or fixed length, etc), and electronics (logic for grabbing single items, small stacks (vanilla normal), large stacks (vanilla stack), whether filter inserter or not) which all must be combined, maybe with some restrictions like module slots so that every range needs a module, filters are only 1 different item per module, etc .... a very complicated system and probably not woth it, but fun to think of :-)