From a reliability/fun factor point of view, the closer you get to the engineered minimums in any capacity, the more likely you are to have the probabilities work against you. Said differently, the closer you get to the smallest possible radius for anyone scale model, the more problems you will encounter. Usually that means more snagging, dragging, or even derailments. If it doesn't happen when trailing the rolling stock, it will happen when you shove them because the slack in the draft gear has been bunched during the shoving.
Often it isn't that the trucks and wheels can't deal with the curvature, but the couplers between the cars, or details near the ends of the car frames, abut each other as they turn, and you get derailments.
The cheapest and best use of your time when you are intent upon testing the limits is to actually measure them. Use the real items, including the tight curves you can mock up temporarily on a board, power the rails with some alligator clips and feeder wires, and see for yourself what is possible, and when you get too many problems for your taste. With that firm knowledge, empirically derived, you can confidently duplicate the same thing on your layout. Note, though, that when you add curves, kinks, turnouts, and grades other than what you tested, all bets are off.