I am a HUGE advocate of properly constructed simulation to train and test pilots, and provide for each pilot, the confidence that they need to have to do the job we all expect them to do.
During in-flight (not simulator) training on the 737-200 a long time ago, we were given simulated engine failures after V1 by closing one thrust lever to idle and completing a circuit and asymmetric landing or a go-around. The FD was the Collins FD 108 which we switched off for the purpose of the exercise. The engine failure climb out was quite easy to handle.
I found this in stark contrast to later versions of the 737 (Classics) where the aircraft was equipped with the now almost universal "twin needles" FD. Despite many hours in 737 Classics simulators, I have always experienced great difficulty trying to "fly" the FD needles at instant of engine failure and initial climb to flap retract height. Although well aware of the folly of `chasing` the FD needles during engine failure climb out, I find it well nigh impossible to "look behind" the FD at the aircraft attitude. In fact I do much better if I glance over and fly the miniature standby ADI, which is unencumbered by needles, to maintain an acceptable climb out attitude.
While this thread is about crosswinds operation, it has also developed into most interesting discussions on simulator fidelity in general. Hence my comment on FD use in initial one engine climb. Could this be a simulator FD fidelity issue? Are the needles too sensitive to the slightest yaw or roll thus causing over-reaction by the pilot? If I can do a perfectly accurate raw data engine failure at V1 and initial climb out without over-controlling, why is it that with FD operating I am all over the sky trying to pin the needles to the centre? Frankly, I must admit the FD spooks me with engine failure and I suspect a lot of other pilots too. In my book the FD is designed to be an aid - not a hindrance.