What usually happens when running on kerbs is that grip fluctuates rapidly causing oscillations in road-load which can set off driveline resonances (which tend to be quite low frequency).
The important aspect of this kind of torisonal response is that it is not caused by engine torque fluctuations and therefore can't easily be replicated on a dyno (unless it is measured and then "played-back" which is not a predictive method i.e. can only be replicated once it has happened on-track).