Why do preheat fluorescents (in Canada and US) usually use choke ballasts over HX or CWA ballasts?
What is really inside of a magnetic rapid start ballast?

The ballast have to supply voltage high enough to restrike the hot lamp on every cycle. In 120V countries mains voltage is low (compared to the lamps arc voltages) which limit the max. lamp for which the mains voltage as OCV is sufficient

For the shorter lamps where the mains voltage is all what's needed, the ballast have to just set the current - a series inductor does this

For the longer lamps the OCV have to be boosted with a step up transformer. So, you could boost the 120V to 240V (or any other volage you'd like...) with a plain transformer, then use a simple inductor ballast on the 240V

But the ballast is a coil itself - can this coil be made to function as the secondary of the step up transformer and as the ballast at the same time ? This would eliminate the detached ballast completely leaving in just the transformer, which simplifies stuff

Well it can. If we limit the amount of power that can go through this transformer, this will limit the power the lamp gets. This can be done by playing with the amount of magnetic fux that can go in the secondary winding in the core, by making it's coupling with the 1st winding loose intentionally. This is what's going on in HX



Now this is what we need to keep the lamp working, but how to start it ? there are many ways :

Heat up the electrodes, then apply a kick of voltage = Preheat

Heat up the electrodes continuously untill they can start a discharge at the existing voltage = Rapid Start

Just give a high voltage immediately = Instant Start



Either way of starting can be combined with either way of ballasting. Examples :

Series inductor + Preheat = the standard Preheat circuit for 20W fluorescents

HX + Preheat = preheat circuit for long fluorescents that need higher OCV

Series inductor + Rapid Start = some "rapid start" electronic starters

HX + Rapid Start = standard Rapid Start

Series inductor + Instant Start = Standard Mercury Vapor circuit in 240V

HX + Instant Start = Instant Start fluorescent



Preheat is done by shorting across the lamp momentarily (either with manual switch, relay, or starter) so the current heats the filaments in series. Then the circuit is opened and high voltage kick is generated by the ballast

Rapid Start is done by supplying voltage to the filaments all the time. The HX ballast contains 2 extra windings on the transformer which supply low voltage to the filaments (a different hack is used in electronic starters but with same result)

Instant Start is done by supplying high enough voltage. The voltage is sufficient to start the cold lamp as is, then the electrodes heat up in the hot spots (and become hot cathodes) from the heat of the discharge



When mains voltage as OCV is sufficient, the series inductor ballast is used. Preheat is the only starting way that does not require any additional complication in the ballast, so is preferred. For Rapid Start you'd have to have a step down transformer just to heat the filaments, and for Instant Start you'd need a step up transformer to get the high OCV

When mains voltage OCV is not sufficient, you need a transformer anyway. So adding 2 more outputs to the existing transformer for Rapid Start is no big deal. And even simpler, you can just make the escondary coil large enough to have enough voltage for Instant Start

So in  HX Rapid Start we have 4 coils :
Input - a coil across 120V
Output - a coil with high OCV but limited power output capability, as it is partially magnetically separated from the input coil. This is connected to 1 wire on either lamp end
Heating - 2 coils that just get few V all the time. They are connected to 2 pins on either lamp end

Why do preheat fluorescents (in Canada and US) usually use choke ballasts over HX or CWA ballasts?

Not only US and Canada.
But the reasoning is a bit in opposite way: Series choke is used, whenever the mains voltage is sufficient as OCV for given lamp, because it is the most efficient, lightweight and so cheapest ballasting topology (the efficiency except for very low arc voltages, relative to the used mains).
As these ballast have only the mains as OCV and no means to provide heating supply other then bypassing the lamp, the starter is required to provide alternatively both the heating circuit and the ignition high voltage.
But it is limited to arc voltages in the range about 60% of the mains voltage, what mean only low power lamps in the 120V areas (so US, CA). Higher power lamps need there a step up transformer (CWA, HX,...). But when there should be already the burden with an expensive (in all: purchase cost, weight and losses) transformer, other, more comfortable systems could be made without any extra significant cost (actually this extra is frequentlly cheaper then the starter with it's socket and it offer more flexibility), so such ballasts were rather designed as RS or some even IS.

In 230V areas vast majority of used lamps have arc voltage below 100..130V, so may work with the simple series choke, so you almost do not find any other magnetic ballast then series choke preheat in the 230V areas. For some lamps were made as the RS ballasts, as these lamps suffice with the mains voltage for hot starting.

What is really inside of a magnetic rapid start ballast?

There are generally two basic forms of RS:
"120V" (not limited for 120V, but usually at least offering this mans level - mainly US and CA) contain a mains fed (auto-)transformer with two separate 3.6V windings for filaments and one for the lamp arc circuit. The 3.6V windings feed the electrode heaters (filaments) by a constant voltage, while the lamp circuit secondary is made so, it 's OCV ignite the hot lamp and it's output impedance (made either as leak transformer or series capacitor) feed the designed current into the operating lamp. As the filaments are fed by voltage, their temperature remain more-less constant (the tungsten resistance thermal coefficient help a lot here), regardless of the exact arc current. This allow to feed the lamp with virtually any current (up to the rated one) without causing the operation with too cold electrodes (so accelerated sputter), so allow e.g. powering regular, cheap 40W rated lamp on 25W only, when lower brightness is required (in homes,...) As F40T12 were replaced by F32T8 in the professional sector, more optimized 25W lamps start to appear more and more often for the home use (they save about 1W in filament heating power, are thermally optimized, so could reach higher efficacy and fall below the power limit, where is mandated higher system efficacy, out of reach for magnetic ballasts with T12 lamps)

The second system use current to the PFC capacitor to heat up the filaments, while the mains voltage ignite the lamp (European SRS).
This is accomplished by splitting the series choke ballast winding to two thinner wires. One feed the system from the mains to the lamp, second to the PFC capacitor. When the lamp is not ignited, the current flow through one winding  to the filament, then through the second wire to the PFC capacitor and then through the second filament to the neutral. For this current path the magnetic field from both winding cancel out, so the ballast provide only small impedance (only small inductance due to not perfect coupling between the windings). This small inductance, together with the PFC capacitor boost the voltage a bit (by the Ferranti effect), what ease the lamp ignition.
When the lamp ignite, it form quite low impedance. Now the arc current come from two ways: Part from the capacitor (the reactive force) through one winding and second part from the mains (the real power plus some of the reactive force). Both currents are not as much off-phase, so they load both windings in very similar way as it would be single wire, almost reaching the efficiency and weight of the simple series choke, while allowing to eliminate the starter. The only costs are the need for two windings (a bit more processing cost) and a lamp sufficing with only slightly elevated mains voltage for hot ignition (so do not work on Krypton filled T8 lamps). So as the T8 ruled the fluorescent market in Europe, the SRS topology fell out of interest (on account of classic preheaters capable to operate both T12, as well as and newer krypton T8 - unlike in the US, in Europe the T8 were designed to be fully compatible with the glow starter preheat ballasts already in use)

Unfortunately the T8s we were forced to use were pretty useless at the ambient temperatures in which T12s worked fine - struggling to start and taking an age to finally warm up, if they ever did. Modern T8s ARE better in that respect but I've seen plenty of installations where the tubes are shimmering and flickering, never reaching optimum temperature, so any potential improvement in efficiency is wiped out.

I saw one fitting which had T12s installed and this was MUCH brighter than the surrounding ones with T8s in.


BG

@BG101: This is the reason, why the T12 should not be banned and why people should use their brain and not rely on promotional statements like "Cheapest to run,...".
The thing is, then majority of the energy consumption of the 4' installations are in offices, where the temperature is within quite tight limits, so where the T8 problems do not show up, so these installations may benefit from the higher T8 efficacy. And that made them the highest volume type, therefore the cheapest ones.
Then someone buy the cheapest fixture with reasoning "it is only fr the barn, so I do not need anything special". But oops. In the barn could be pretty cold (or hot) and the T8 stop working at all.