There is no electronic ballast, because the standard, as it is today, can not guarantee really all such lamps will work.
The main problem with present lamp designs is the lack of the capacitor specification in the standard.
But the standard even does not specify the type of starter used there - someone may theoretically make his lamp (still complying with the standard) with an electronic "pulsestarter" - and that won't work with any HF ballast at all.

To make these lamp HF compatible, the lamp specification would have to really restrict the lamp design to glowbottle starter parallel with specified capacitor. It could be written so, that practically all present designs (except the cheepeese with the capacitor missing) would meet that requirement (the capacitor specification could even be just some range, not really exact value), but for that to happen, at least the major players would have to agree on such restriction.

The glowbottle starter can not be used to start the lamp, as with HF ballast the inductance is too low for just an inductive kick high and long enough to ignite the arc, the only option is really the resonance. The starter really can not function there as the preheat and starting switch, the way how the glowbottles work may be made to either do not interfere with the HF ballast at all, or serve just as a lamp EOL ballast protection, there is nothing else the glowbottle could do.

The "pulsestarter"-like, or all other electronic starter concepts I have seen would even mean the HF ballast won't just work at all.

The standard really allows just virtually any starting device operating with Henry-range inductor (that is, what leads to the 0.17A for these lamps) in series and 50/60Hz operation. That is too much freedom for the lamp maker, restricting the ballast maker to just the series inductor ballast.

Of course, someone may sell a HF ballast+lamp combo, but that is then rather proprietary combo, not complying to any standard.

What about Instant Start-like ballast - Provide a kick from the outside right away, not caring for the starter at all ? (he capacitor for the resonance is in the ballast parallel to the output)

What about Instant Start-like ballast - Provide a kick from the outside right away, not caring for the starter at all ? (he capacitor for the resonance is in the ballast parallel to the output)

Even the real instant start (e.g. the US F32T8 with single wire per lamp end) do not provide single kick, but they as well use the resonance to boost the voltage.

Then the eventual capacitor presence would be problematic - it will connect parallel to the resonant one within the ballast and in that way alter the parameters of the output circuit.

Make the one in the ballast >> the ~5n used in the lamp, so it won't matter much

Having it "so the 5nF would be negligible" would mean at least 15nF. And that is way too much.
The mentioned 5nF is already on the upper edge for the 0.17A lamp (in fact the typical values are rather around 2.2 or 3.3nF), higher capacitance would mean a necessity to lower operating frequency bellow 20kHz. And that would mean larger magnetic in the ballast, so higher losses there, higher stress for the already "on the edge" electrolytic (the lower frequency is tolerated way worse than the higher frequency current) and whining ballast.
Plus the larger total capacitance parallel to the lamp would cause arc stability problems (that is the main reason, why the capacitors with the glow bottle starters are just the few nF and not larger), yielding flicker and faster wear.

As i understand the Frequency of the ballast is proportional to sqrt(LC). Make lower L to keep same frequency ?

As i understand the Frequency of the ballast is proportional to sqrt(LC). Make lower L to keep same frequency ?

That is just one part of the equation.
The other part is to reach the required ballasting impedance, so the ~150VAC from the halfbridge get translated into the required 0.17A into a 70V arc lamp. So that means:
L = sqrt(150^2-70^2)/0.17/2/pi/Freq
So if you put those two equations together, doubling the capacitor means halving the frequency and doubling the inductance.

Why the voltage have to be the same ? Regulate that to match the new L....

First of all if you have an electronic ballast you don't need a glowbottle it serves no purpose the whole point of electronic ballasts is that the use of HF means the tube self starts after a short heating time secondly EOL detection should be easy to do just need to detect a higher impedance or wrong SWR and get it to shut down is easy. Third point I saw some tube with patterning on a bus recently the stritations seemed also to be affected by the engine revs which makes me think maybe drive voltage plays a part what do you think?

For the PL ballast many thigns ca be done for EOL protection, but for now lets just get it straigth how we run it on HF at all.....

In the bus, the voltage in the system, as well as frequency of ripple present in the system (rectified output of alternator) change with engine RPM. The voltage affects the fewquency and power of the HF ballasts (the ripple in the bus system is at frequency << the HF, so it can be treated as momentary DC voltages when ve evaluate the reaction of the HF ballast to it)

@Hannah: The point was how to create an electronic ballalst for a PL-S lamp, where the starter has to be in the socket to work on the magnetic choke.

@Ash: The voltage is dictated by the mains - the 150V is the rectified 230V minus half of the tank capacitor ripple (so about 300V in average) divided by two (the inverter is a halfbridge, so it's AC output is always half of it's DC supply).
To have the DC regulated to other voltage would mean an extra regulating stage in the ballast, no one would do that.

Why not ? How complicated can it be ? And what about controlling the width of pulses from the inverter (feeding into L) ?

1) Extra cost
2) Extra size
3) Extra losses
4) Enormous reactive power for starting (the 2.2nF designs already operate at 50..100VAr in the resonant tank during ignition, using 15nF instead would mean higher 100's VAr), just because of too low characteristic impedance of the LC tank (Zo = sqrt(L/C) and it characterizes, how high currents would have to flow in the circuit for the given resonance voltage; normally the design is made so, it provides reasonable heating current)

....Really that much extra cost/size/weight to forbid the building of the ballast ? I think the PWM option can be implemented without adding expensive "raw material" sort of component (such as big inductor or capacitor), just a control IC w/ current sense instead of the toroid feedback to the switches.....

Same for the preheating, why let the LC characteristic rule it instead of PWM control ?

Just the PWM alone does not regulate anything, you always need with it a loss-less filter to turn it into an efficient regulator. And that loss-less filter may be e.g. the intrinsic motor inductance in the case of regulating a DC motor, a thermal inertia in the case of a heater regulator, or an explicit LC filter in the case of an electric power converter. Without LC it just does not work at all.

The fluorescents are fed by AC, so all the conversions are mainly AC impedance or resonance controlled...

The question is not only for the preheating, but for the HV generation. If you have a 15nF capacitor and want to generate 600V across it at 30kHz resonant frequency (so the steady state operating may be the 20kHz), the reactive power would simply be about 200VA (with 2.2nF at higher frequency it is about 70VA).