It's generally accepted that a constant, steady platter speed will produce the most accurate reproduction. Above, Chris stated that a high mass platter will produce speed variances of low magnitude and long duration, and that light platters will produce speed variances of high magnitude and short duration.
A main goal of TT design is steady platter rotation, but could TT designers be barking up the wrong tree? Shouldn't we consider the speed variances of the cutting lathe and try to produce a TT whose platter mass, motor power, and stylus drag combine to approximate the speed variances of the cutting lathe? If the cutting lathe's platter slows down and recovers due to a heavy modulation, it might not sound right when replayed in your living room if your TT's platter also doesn't slow down and recover at the same rate. Is it possible that rock solid speed stability in TTs is actually a bad thing?