You are up against laws of physics as Ravenser wrote. But despair ye not, because railway modelling is in part about engineering, much of which is about understanding the problem, then finding a way of overcoming or mitigating its' effect.
The Back-EMF facility found in both DC feedback controllers, and the better DCC decoders, offers at least part of the solution. This technique attempts to keep the motor turning at constant speed when the load varies. If that does the job then you are sorted.
However, if the load behind the loco is enough that the driven wheels slip badly when on the curve you may still not maintain speed. To overcome that requires more weight on the driven wheels. That will usually be enough to do the job.
If speed is still not maintained then either the power supply system or the motor capability may not be up to the job. In which case operate long enough applying the stress, and one of the components will fail. Replace failure with a higher grade component (this is a very respectable technique, and was the principal basis of leading edge aviation development until relatively recently!).
