I very much doubt noisy brushed motors driven by PWM will be used.
Brushed? Where did you pull that out from? You're the only one who has mentioned anything to do with brushed motors here.
I'm starting to think you know nothing about this, and are just spouting your gut feelings as if they were facts.
PC fans were always BLDC, with the PWM frequency between 21 kHz and 28 kHz (25 kHz typical). RPMs vary between 300 and 3000, i.e. between 5 and 50 rotations per second (although smaller, 40mm fans as used in servers can be significantly faster – and horribly loud with a whiny sound spectra).
The cheapest way to implement a small PWM-controlled BLDC fan is to drive the coils using a half bridges (with BJTs or FETs) directly modulated by the PWM. This is possible, because the duration each coil is powered is many PWM cycles long, even at the highest speeds. Especially at low PWM duty cycles (thus low RPMs), this causes a noticeable characteristic whine. It is more noticeable with larger fans (120mm and larger), unless they use better driving schemes like e.g. Noctua uses – either proper filtering, or using a proper BLDC driver that only uses the PWM as a control input and not directly to modulate the output coils. It was such a problem that all but the cheapest manufacturers have had to shift to better filtering and driving schemes, because the cheap ones were rejected by the customers. Today, if you open up any but the very cheapest PC PWM fans, you'll find a dedicated IC and quite a few components in the hub PCB.
Ceiling fans rotate at very low RPMs, say 30 - 300, and with their large blades, have quite heavy fan blade assemblies. Thus, it will be in the interests of the cost-cutting manufacturers to rely on the blade inertia and coil inductance for the filtering, and that
will lead to audible motor noise. (I am not talking about levels audible in an office environment with idiots blathering on the phone nearby, I'm talking about the important situation: using a ceiling fan to help one sleep better.)
I am absolutely certain there are many members on this forum who will
insist that the inertia and coil inductance
is sufficient filtering. It is demonstrably not so to me.
Compare to e.g. TI TPS61165 and TPS61169 high-current boost LED drivers with 5 kHz - 100 kHz PWM dimming control. They do not use the input PWM signal to modulate ("burst") the LED current either, because it causes audible noise in the output capacitors at (multiples of) the PWM frequency.
Is this just marketing wank? No, those who do care know about the difference it makes in quiet environments. Do all manufacturers use such techniques? Definitely not. Many of the cheaper 12V (CV) or 350mA (CC) LED drivers do emit audible (to me) noise around 20 kHz, say between 10 kHz and 25 kHz, especially so with low duty cycles (load on the lower end of the power the driver is specced for). I do not use those, either, which means I can only buy from places that let me return the item if they do produce such a sound.
Thus far, audible noise, especially near-ultrasound noise, seems to be something completely ignored by policymakers. They only consider the acoustic power, and not the audio spectra. Yet, white noise is soothing, but a pure sine wave at the same loudness can drive a person crazy.
I wonder if it is because their own hearing is utter shit, or whether they just don't understand things that affect themselves.