Basically you want at least acceptable efficiency and high power output from a motor(and a reasonable power spectrum over a wide spectrum of angular velocities, reliability and longevity): Once you have those factors figured out, it's only a matter of appropriate gear reduction(which could possibly be implemented through sprockets, traction belts, etc., depending on the application.) to adjust towards your speed requirements so that you'll have to appropriate torque, etc. If you have some sort of variable gear reduction(like that of a transmission.), then you can have an even wider power spectrum over the angular velocities which means: High power at low speeds and high speeds! Things are going to be fast now!
The only problem is with the implementation. DrunkenSkunk(sp? Or was it DrunkSkunk?) brought up the point that the bearings wear out from some such high speeds of RC motors and the high speeds bring about a heat issue, which degrades efficiency and performance. There's also the issue of the gear reduction ratios needed for e-bike applications and that has its own brand of difficulties. And then there's the transmission possibility, etc.
It's clear a "middle ground" would be more effective. A lower RPM, higher torque motor with the same demands of power but still lighter than the crystalyte type of motors; It'd be heavier, yes, but it wouldn't wear the parts out as quickly through its heat problems and wouldn't require absurd gear reduction ratios. There'd still be the issue of implementing variable gear reduction.
But, is there really a motor-middle ground? Something with the power of a traditional RC motor and an insignificant weight penalty, but with the higher torque and lower RPM more fitting for an e-bike application? Methinks yes, there *has* to be given the "continuous"(as opposed to discrete) nature of motors and motor types! But I'm just too ignorant to know what ideal motors fit the bill.
But now it's starting to sound like geared-motors may have already solved this. hmmmmmm...