John I don't agree 100% with your line of thinking. There's nothing inherently wrong with a high-ratio gearbox.
In NMRA modelling contests, to get full points a loco has to run smoothly at 4 scale mph (which is also the maximum coupling speed of the prototype.) If the drive "leaps" from zero to 4 mph, that greatly detracts from the observed realism. But the 4 mph number is a good starting point.
My experience testing many different locos in O gauge over the years, is that common can motors, including Pittman motors, just don't run consistently below about 800 RPM when the loco is running "light." Depending on how quickly current consumption increases near stall, and the characteristics of the drivetrain, it might run slower for short distances. But a drive designer shouldn't bank on that.
With 73" drivers, to achieve 4 scale mph at 800 RPM, you need -- surprise! -- 44:1. Samhongsa knew this, because they built many brass locos in other scales before they took over and improved the Williams Crown Edition line in the late '80s. That's why those Williams gearboxes say SAMtech.
Personally I think the top speed, or even the typical speed of the prototype is irrelevant. On a small layout with typical 3-rail curves, 60 mph or even 45-50 mph is plenty fast. So if 800 RPM = 4 mph, you'll need at least a 9,000 RPM motor to reach track speed. This has been the reality in all other scales, for years. If you need a bigger power supply, or special attenuation of NVH for high-RPM operation, that amounts to built-in quality of design.
This is the direction 3-rail O scale was headed too, until MTH entered the market circa 1992. Coming from a tinplate background, their designs were a throwback to 1950s toy trains. Tall gear ratios. No separate gearbox or "bottom plate" with removable wheels & axles. Only small-drivered models (like the Y6b) were capable of smooth operation at switching speeds. This was definitely a step backward!
Speed control came along in 2001, introducing a new set of possibilities (and headaches!) Like a giant band-aid, it greatly masks what I perceive as a tragic design error. To me, 30:1, even 36:1 with a 10,000-12,000 RPM motor would be ideal. With a balanced drivetrain in a die-cast metal body, that shouldn't pose insurmountable problems with NVH. This combo would maintain a wide range of realistic speeds without cruise control, and run phenomenally with it.
The command setup is icing on the cake, and it should NEVER be the constraining factor. If there isn't a control system and decoder combo on the market that can handle a 12000 rpm motor with enough torque for O scale, then someone needs to invent one, pronto. Because ^^this is the right way to build a train!
