Fetcher is close to the point.
Adding the freewheel mech is going to introduce complexity and added weight, size etc.
There's a way around this by removing the stator tooth completely and lay the winding in epoxy directly on top of the iron between the magnet ring with its vector aligned to the magnets, and use the inductive coupling to the back iron to dictate the gap between the magnets and iron size vs the operational voltage and torque this will give the size of your total available copperfill and motor size.
What you will end up with is something that looks like a slotless motor that has no cogging forces at all and can still produce high torque without breaking freewheel pawls and will have less saturation of the iron so it's capable of high rpm if the magnet ring is built with gaps but you will lose initial torque due to being able to operate at high rpm without cooking the magnets.
I've been researching about slotless motors and how magnets suffer with eddy currents in my spare time so this is something I can see that could be improved. But you will lower torque output compared to the cogging version no doubt with like for like material content as the magnets have no gap's between eachother and are compacted tight, that's bad for eddy leakage between each magnet at increased rpm that's why magnets fry at high rpm if the ring ain't got eddy gaps or magnet segmentation the leakage increases with rpm but for stall torque it's going to give a increased torque figure.
But at the end of the day we measure a motor by its total output power so it's rpm handling is a big deal and that cogging torque is robbing the motor of its overall torque output as it has to be overcome before the motor will turn and throughout the full rotation it keeps happening with every phase pulse there's a cogging torque loss.
Lots of good reads and info at this site how commutation works lots of eye openers.
https://www.motioncontroltips.com/whats ... ss-motors/