Couple of thoughts, where we are looking at new chums transitioning to jets.
I always found it very useful, for such folk in the initial OEI work, to run a couple of failures on the longest sim runway I had available but get the pilot to stay on the runway and just rudder the aircraft down the centreline.
Providing the box fidelity was half sensible for rudder (prior to the FAA's push post the 737 hardover problems this wasn't necessarily the case for -200 sims) this provided a good idea of the relationship between speed and rudder input required to keep on the line after the initial divergence was brought under control.
The second exercise, as competence improved, was to spend some time progressively reducing to a Vmca/Vmcg limiting schedule and CG. The arms and legs exercise required resulted in no further problems with "normal" failure sequences ....
The better manipulators regularly ended up able to handle the worst the box could throw at them for a 0/0 takeoff with a requirement to backtrack the LLZ. The effect on confidence was dramatic.