This is a difficult issue and there is probably an optimum for each magnet size and spacing. If things were decided on volts alone it would be fairly easy.
Any turns smaller than the coil are not cutting all the lines of flux and are not contributing full volts, but they still contribute something and they add relatively little resistance. I suspect you are justified with a round magnet to reduce the width of the hole considerably, I am not so inclined to agree about reducing the height, you may be better off with oval coils with the long dimension equal to the magnet diameter.
With rectangular magnets I would keep the height of the coil hole equal to the magnet. I would also keep the outer width about magnet width or not much less. As the magnets are close on the inner diameter you may possibly bring the width down to half magnet width where there is more leakage flux. Dan seems to get good results this way.
With larger magnets I would keep more nearly to the hole being the size of the magnets. Small magnets have quite an area round the edge where flux is well down on the centre value, that may be why the small round hole comes out well in this case.
," i would have to say no ,(see next paragraph) thicker coils do not give more power ,because if you have the voltage you need anything more is just added resistance."
Now this is where I am beginning to disagree. If you have got the volts you need and have space to spare, you use thicker wire. Maybe I misunderstood and you are trying to optimise for best output with a given wire size.
The way I see it is that you need 2 mechanical gaps and with small air gaps these waste a significant percentage of the winding area. With a 2mm thick coil you would waste about half the space.
By using a larger air gap you have less flux density so you need more turns, but you have increased winding area to use more turns of thicker wire and still get the volts with less resistance. Neglecting leakage flux, the magnet works at its best point when the air gap is equal to total magnet length (about .6T)
In real life you reach the .6T point with the gap less than total magnet length.
For a dual rotor with 2 magnets in series and with average spacing and leakage this generally means that the best coil thickness is about the same as one magnet. That leaves room for fibreglass and air gaps and will work in an air gap about 1.4 times one magnet length.
Still room for experiment.
Flux