"As the COTS units are "super cheap", one has to assume you want to do something MORE -- or differently..."
He can *buy* the "ultrareliable reed switch and magnet". Presumably, that doesn't meet his needs. It doesn't, for example, register correctly if the magnet happens to be CLOSE (rotationally) to the switch and you are stopped -- chatting with someone -- while rocking forwards and backwards with the bicycle. (it has no sense of direction)
It doesn't register variations in speed within a single revolution (e.g., cranking up hill) or the instantaneous WHEEL acceleration from the rider's efforts.
It also requires the wheel to be true (no wobble) to ensure the distance from magnet to switch is sufficient to reliably engage -- while not getting
*whacked* by some other portion of the wheel as it passes that isn't true.The magnet can be dislodged, sensor alignment botched, etc. If you swap out the wheel, you have to remount a new magnet on the replacement wheel, ensure proper alignment, etc.
And, you're still going to have to run a (electrical) cable up to some sort of display/control unit. And, protect that input to the electronics (i.e., you wouldn't want to directly expose a pin to that external signal) lest the device die with the first bit of ESD.
Instead, bring the motion (ALL of it) up to that unit and let *it* decide what criteria are appropriate to measure. Instead of cable-in-cable plugging into the back of a legacy "speedometer" (indicator), let it drive a slotted wheel/gear mounted on a board already aligned with the sensors (which need not worry about the same environmental issues that sensing ON the wheel imposes) and UNBUFFERED inputs to whatever processing/display/control logic.
Most of his designs worked!