3 pole BLDCs work at any % > 0% and always start without dithering. The logic is between the Hall sensors and the drivers.
2 pole BLDCs used in muffin fans stop at the top dead centre (TDC) midpoint cog position of permanent magnet torque. The Hall sensors must lag this TDC otherwise by a small amount to always have a positive direction for torque. They allow the minimum lag for manufacturing tolerances. This is why 2 pole fans sometimes fail to start unless given a small nudge. But it has nothing to do with 100% PWM but does illustrate the difference between 2 and 3-pole BLDC motor's torque profile.
The commutation logic works externally with the logic levels of the Hall sensors. In 2 different companies where I used high-volume BLDC muffin fans, I experienced the same fan failure in a small % of the population. I designed a 30-second pulse start-stop tester for 10 fans in parallel for incoming inspection to allow the fan to stop at all 4 stop positions. This was necessary because 99 % of these fans worked from every resting position but only 1% failed usually only in 1 of 4 Hall positions which would cause it to fail to start or just dither back and forth fast. I disqualified the vendor Nidec (large expert fan co.) who accepted my test design, took over the task and immediately corrected their sensor location margin problem on future shipments.
I hope you remember this if you ever see a 2 pole BLDC fan need a nudge to start but a 3 ph fan will never behave like this even with 100% PWM