This issue is a mess. You can continue squeezing magnets in until they touch and you will increase output, but you will be making very poor use of the magnets.
It is ok adding more magnets but unless the coil can link the total flux for that magnet you get poor value for money. Yes add more magnets, that increases total flux. Yes it does increase frequency and that increases output. To be really useful you must increase the diameter if you add more magnets. If you have magnets touching ( or virtually so) you can't effectively use any width of coil leg ( no winding space)
For conventional 3 phase windings in slotted cores the best spacing is roughly with magnets about twice as wide as the space between them.
For the odd part wound 12/9 arrangement you do better with more space between magnets and space at least magnet width is desirable. With radials this is a compromise and space about magnet width about half way out along the magnet seems ok.
If the loop within the smallest turns of a coil is smaller than the magnet then you don't link all the flux. Because there is a trade off between induced volts and resistance it seems justifiable to make the hole in the coil smaller than the magnet but if you end up with magnets touching then turns that should be surrounding one magnet will be linking flux from an opposite magnet. You can effectively only use coils with legs one wire wide and that gives no winding space. This is much more of a factor than the issue of leakage flux between close poles ( true only for neo).
Within reason the coil shape is of no importance as long as you link all the flux.
Contrary to what some still want to believe the length of the coil is of no importance. If you make it excessively long and thin you end up with low values of area for a given length of wire. Volts are ok but power will be less.
Commercially high speed alternators are long and thin with few poles, low speed ones are short and fat with many poles. Most of this is linked with the mechanical constraints. Windmill alternators are best short and large diameter but not with large numbers of long thin poles with little gap.
Long thin poles are probably a pain with axials more so than radials.
Flux