Is it possible that the motor isn't in the bike, but rather on the rider? For example, could something be fashioned from an electroactive polymer that is sufficiently thin and flexible that it can be hidden under the bike shorts and, with electrical stimulus, help drive the upper legs down (or help pull them up)? If so, using such a device would possibly not require collaboration with the team (at least not to the same extent or with greater prospects for credible deniability from the team), and wouldn't be detected if the bike and the rider's bodily fluids are the only two things being tested.
I can imagine that using such a device might take some practice, especially to synchronise the stimulation with the gearing on the bike. Also, in order not to injure/bruise the rider, it might be better to have it work at lower levels of force, but higher frequency. Could such a device explain a donkey-to-racehorse transformation (which even took the rider's team by surprise), egg-beater cadence, occasional yo-yoing and crashing going uphill (i.e. synchronisation problems), etc.?