During my 737 3-400 type rating training I clearly recall a demonstration in which stab trim was allowed to go beyond the point where manual trim could be operated due to control loads. Whether this was part of a stab trim runaway drill or an extension of demonstrating the trim system I no longer recall. The purpose was to enable awareness of the condition and practice the technique of (both pilots) hauling back on the yoke and then relaxing the pull while simultaneously clawing back a little manual trim using the wheel. This has to be repeated several times before the loads reduce enough to allow continuous trimming without unloading pull on the yoke.
I was trained in the 737-200 simulator in 1977 on the technique you described. One rather crude method of demonstrating the roller coaster sequence (and I emphasize this was purely 'playing around' and not a structured planned exercise so please no criticism) was to get airborne with the stabiliser trim purposefully set full electrical forward to see what stick force was needed to get airborne.
It required full back control column to get airborne as can imagine and we passed VR plus about 20 knots before the nose would rise. The purpose was to gain enough altitude before attempting the roller coaster method of relieving aerodynamic forces on the stabiliser. With the body angle pulled hard to 10 degrees nose up, or more if possible, against the full forward stab trim, the pull force was momentarily relaxed and the aircraft 'bunts' over (for want of a better word).
During that very short period that the aircraft is transitioning from nose high to nose low the manual stab trim wheel was able to be very rapidly wound back from full forward for about five seconds before the nose dropped and height was lost.
Both pilots then again pulled back hard against the aerodynamic force until the nose was high enough to repeat the exercise. As soon as the pull was relaxed and the aerodynamic load momentarily eased, more frantic winding back of the stab trim handles was available for another five second until the nose dropped again. It is emphasized that it is extremely hard to wind the stab trim handles back unless the aerodynamic load against the stabiliser is reduced
From this rather amateur experiment it was clear that below about 1500 feet the success of the roller coaster method of using manual stab trim was very much dependent on how high the nose could be raised above the horizon before the elevator could do no more and backward pressure let go to enable the manual stab trim to be operated relatively freely for that few vital seconds you had before pulling back on the elevator to stop the ensuing dive.
Two pilots using their combined strength was needed to keep the nose from dropping but as soon as the PF took one hand from the control wheel to wind the manual stab trim, it was difficult for the other pilot to stop the nose from dropping sharply.
Once you have seen this quite dramatic exercise demonstrated in the simulator you will never forget how to fall back if necessary on the roller coaster method of using the manual stabiliser trim to relieve the aerodynamic forces on the stabiliser.