My 2x f40t12 hpf rs electronic ballast produces a thump when started in the morning. I'm not sure what causes it but it's cool feeling the whole fixture vibrate when it happens.

https://www.youtube.com/watch?v=hHT6MptrhIU

Running a GTE Sylvania grolux and a Philips Alto daylight

Isn't that the type that used a series choke on the mains input (to keep the input current harmonic content under control)? Then the inrush current when the DC filter capacitor get chsrged on the power ON may create such noise, mainly when some not that rigid magnetic metal sheet is nearby..

  What?

There is a picture of the ballast below.
I don't think it's the stuff around the ballast since it doesn't fit the desc. of non rigid magnetic metal.

Something like this.

It is how the electronic ballasts work: The mains get rectified, the resulting DC smoothened by a large capacitor and then fed to a 2-transistor chopper circuit, which then converts it into HF AC voltage. That HF voltgage us then fed via a ballasting choke to the lamp. Plus there are resonant capacitors, which make out of ot an LC resonator tank, which boosts the voltage for ignition.

Now from the input perspective, it is a rectifier with a capacitive filter. This then consumes current only when the AC input voltage is around its peak, so all power is concentrated into rather narrow, therefore high current peaks. These peaks contain a lot of harmonics of the mains frequency, which are loading the distribution network (in other words lower the power factor; and cause other problems) but do not actually carry any real power. So it is desirable to get rid of them.
One of tge earliest methods is a filter inductor in series with the input. It "smoothens" the peaks out (by presenting a large impedance towards the higher harmonic currents), so make the input current closer to the ideal sinewave. This method was used in many early ATX computer power sopplies (from late 90's), as well as in some electronic ballasts. Later it was replaced by an active booster power factor corrector, because that was way more compact and less lossy.
And as any other inductor, part of its magnetic field escapes outside of the main core, mainly when the core saturates. And in the power factor correction inductor the core saturates mainly when there flows a huge current, so mainly when the filtr capacitor after the rectifier gets charged at power on. This magnetic field then vibrates nearby magnetic metals (steel, nickel,...), so once these can move a bit (e.g. the fixture steel sheet body which can flex to some extend), it then emits this "thud" noise.

Thanks for the explanation, now I really understand how these things work! If the conditions are right, how strong can a ballast thump be?

What the American calls "Electronic rapid-start ballast" is nothing more than a CFL like ballasts, which works in a similar manner, with or without PTC controlled preheating.

Most cfl ballasts that I've come across don't usually have controlled preheating.  The ballast in the video acts like an instant start ballast most of the time, with less than a second of startup.

These 4wires per lamp ballasts are technically not IS but rather RS, because it warms the electrodes during and after the ignition. Even when it takes fraction of a second.
They apply heating current at the same time with the ignition voltage, plus the time it takes the electrodes to really reach thermionic emission.
It comes from the way the lamp circuit is connected: When there is no arc in the lamp, the ballasting inductor, together with the starting resonant capacitor form a rather high Q series resonant circuit, fed close to the resonance. That means rather high current (usually about 5x the lamp rated current) is flowing through the circuit (so through the filaments) as well as the elevated voltage across the lamp (nearly a kVpp). Once the discharge ignites in the lamp, it forms parallel resistance loading the LC, so reducing the voltage and currents. Once the filaments reach emission, the discharge voltage drop decreases to the normal level (10's till 100V), so presenting resistance so low the resonance is pretty much killed by it. As a consequence, the resonance capacitor current becomes insignificant, so filament does not get virtually any external power anymore, the arc just gets all the current.
Because the heating current is so large (5x the rated lamp current is not unusual), the filament warmup actually takes just fraction of a second.
The high current itself does not blow the filaments up, because all the transitions are in fact controlled by the temperature of the very same filaments, so it never goes into dangerous (for the filament) levels. And it is not the current, but the temperature, what kills the filament when becoming excessive.

Temperature destroys the filament, never knew that.

With my f4t5's, I connect the bulb but doesn't use a starter and I turn on my plasma and put it next to the bulb, the bulb looks like it's being run rapid start where it runs dimly for a second then goes to full brightness. The filaments have the right voltage to them then the plasma gets the arc going then the filaments with power complete the startup.

The dimmer stage is when the electrodes are still too cold to emit electrons, but the discharge is already there. It is the most damaging stage, as there is an extra voltgae drop just on top of the cathode (needed to extract the electrons from the cold surface), accelerating ions that are then "sandblasting" the surface. Plus this extra voltage drop causes the arc current to get reduced (and needing the plasma ball field support), hence the dimmer glow.
These ions carry heat to the electrodes, so if the current is sufficient, after some second or so they heat up the cathodes to the emission temperature.
Once the electrodes warm up so the electrons leave their surface freely, that extra voltage drop disappears, so there is nothing to accelerate the ions anymore, so the damaging stops. And as there is less drop in the circuit, the arc current gets higher (normal), so the lamp brighter.

This also happens with many RS ballasts: Once the filament warms up partially, the discharge starts. But the emission is still not enough, so it is dim. Only when the filaments reach full emission, the lamp turns full brightness.
By the way this is what the "CFL style" electronic ballasts do, it is just all happening way faster than with the "classic RS" (fraction of a second instead of a second or two).

If you would like to really limit starting wear to a minimum, you would have to first heat up the electrodes without any (or very limited, so the discharge can not ignite) voltage across the tube and only once they have reached the temperature (but not overheat it), apply the high voltage for ignition. This is, what preheat would be doing, if the preheating would be long enough. And that is, what the "programmed start" ballasts do.

I've seen a programmed start cfl in my grandmother's basement.

When it's using the plasma ball to startup it does run cold cathode so I don't like to leave it in the 'middle' stage too long.

I wonder how blackening from preheat, programmed start, rapid start and instant start would vary if they were all started 5x a day all exactly the same amount of time from eachother then left on for an exact amount of time. Instant start would probably go out first.

In the 2x54W T5 HO fixture at the kitchen of the low floor of my hostel, when I turn it on, it have this dim stage of rapid-start for a sec, when the electrodes warms up, and don't start instantly. I hope this isn't the reason why the Osram lamps there, don't blackened so fast.

Why don't you hope that isn't the reason the bulbs don't blacken so fast?

When the lamp glow dim, the electrodes has yet to be fully heated, so that glow is a cold cathode mode.

That's why they blackened so fast then, either the ballast is faulty or it's the incorrect ballast which may make the tubes behave like that.