Frank, The total error, i.e. PE plus polar misalignment plus cone error, flexure, refraction etc, must be less that your imaging platform can resolve (in theory) in practice if you keep all of this within your local seeing then it's going to be ok. Remember that PE is just one part of the problem. Take my example, even though I have a fancy shmansy mount with PE < .6" the reality is that after a couple of minutes all of the other factors will cumulate to produce egg shaped stars or trails.
The main reason to tame PE is to make the job of autoguiding easier. No matter how good the PE is you still need to autoguide for very long exposures. As long as the PE is indeed 'periodic' and the rate of change isn't too sudden then it realy doesnt matter in an well autoguided solution. The real problem arises when the error is sudden and happens faster than the guiding can compensate.
  