There have been attempts at the axle-mounted motors approach (google 'Railflyer' for a recently-failed historical example), but you've listed most of the pitfalls, which are very significant. The micro motors have extremely small torque, the associated geartrains occupy a lot of space and require precision engineering, and there needs to be a boxfull of expensive control electronics to attempt to synchronise the motor rotations. For low-power underfloor applications like a DMU, small motors have a good application, but motor shafts need to be mechanically ganged to each other to synchronise the shaft speeds where more than one motor is used. For larger locos, the multi-motor approach becomes needlessly expensive.