Fluorescent electronic ballasts are switching power supplies. They are supposed to work in resonant mode, which is naturally very low in interference emissions

In a very over simplified explanation :

Switching power supply works by switching current to an inductor (primary of a transformer in a PC power supply, ballasting impedance in a ballast, etc) on and off with a MOSFET. The MOSFET is controlled by a signal coming from some control circuitry. It is turning on and off fairly instantly in a square wave form, but if looking closely, it does have loss (turning to heating of the MOSFET) resulting from the power it has to interrupt during the quick transition when it is switched off

Interrupting current through an inductor causes a voltage transient which may be harmful for the circuit, and a spike of interference



Switching power supplies can be made to work in 1 of 2 modes :

 - Hard switching : Simple circuitry is added to mitigate the voltage transient when it is created, and EMI filtering is added to filter the interference. The MOSFET is chosen to be with sufficient ratings to withstand the power it interrupts when it is switched off

Hard switching is used in many low and medium power circuits built to be simple, reliable in a wide range of operating conditions, and fairly efficient, but not trying to be extremely efficient. Typical >80% efficiency is achievable, sometimes more

 - Soft switching : A secondary path through an LC (inductor and capacitor) circuit is provided for the current to continue to flow, in the moment the MOSFET is switched off. The current is allowed to "wind down" in a one-cycle-of-sine wave-like way after the MOSFET is already off. This way the initial transient and interference generated are significantly reduced, and as the MOSFET does not actually in itself brake all the current to zero, it does not have the related losses

Soft switching is used in circuits built to achieve exceptionally high efficiency, typically >90%, in high power circuits where every loss is significant in numbers, etc. In the case of Fluorescent ballasts, there is an additional reason : The lamp is operating in open air and is a very efficient antenna to broadcast any interference the circuit generates



For soft switching to work correctly, some rules must apply to the working conditions of the circuit. Such as the switch on and off timing of the MOSFET. It cannot be switched back on if the current haven't yet completed its "wind down" cycle after the last switch off

If a circuit designed to work in soft switching, is operated outside of its intended working conditions, it will revert to hard switching. As it was not designed for it, it does not have the ability to withstand the transients generated, the minimal EMI filtering present cannot do much with the level of interference, and the MOSFET with moderate power ratings cannot withstand the losses of hard switching. This will lead to increased interference and very possibly to damage to the circuit. (In contrast, a circuit designed in the first place for hard switching, does have everything required to work reliably in the hard switching conditions)

Even when everything else is ok, it may take a cycle (or sometimes few cycles) to establish soft switching after being first started, or in case of abrupt change in load



When operating with an EOL lamp few things may go wrong :



The lamp demands too high power from the power supply (result of rising arc voltage)

When a switching poewr supply is asked to provide higher output voltage, it will prolong the on time and shorten the off time of the MOSFET. If the EOL lamp requires too high voltage, the circuit may shorten the off time to the point where the circuit does not properly "wind down", reverting the circuit to hard switching operation

The overload of the cirucit may trip pulse-by-pulse ("emergency") current limiting protections in the ballast circuitry, which interrupts the current abruptly in each cycle. This will lead to increased interference, and may revert the circuit to hard switching



The lamp conducts current with significant DC component, as one of the cathodes is bad

DC will cause saturation of the ballasting inductor (which is in series with the lamp), leading to abnormal high current draw during each cycle of the switching. The high current causes high magnetic leakage from the inductor which increases audible noise. High enough current may trip pulse-by-pulse current limiting, or damage components if the pulse-by-pulse protection is not present. This will lead to increased interference, and may revert the circuit to hard switching

Most ballast designs do have a capacitor in series with the lamp to block DC and prevent this problem. DC charges this capacitor up. As it charges, it limits the possible output voltage of the ballast, which may make the ballast only marginally be able to continue providing power to the lamp. The lamp may essentially start flickering every other cycle (at HF frequencies, so not visible as flicker to the eye, but below the actual switching frequency, so audible). This may lead to audible noise, and to the ballast being unable to establish a stable soft switching mode



The behavior of the ballast strongly suggests that it can be damaged if you continue running it with this lamp

If you're wondering, this is the ballast: https://www.lighting-gallery.net/gallery/displayimage.php?pos=-251672
I didn't run it for long. Another interesting thing is that when I put a lamp with a busted cathode (it is new but filament got loose), the buzzing noise goes away after a few seconds. So this buzing only happens with EOL lamps, not lamps with broken cathode.

For the most part (this applies also to other electronics), if it is running outside of its intended operating conditions, you cannot know how fast it can be damaged (unless you know that it does have good protections against the specific condition)

The damage may be instant, or it may start accumulating right away (so the ballast still works, but it had sustained internal damage that will remain)



With "Electromechanical" stuff it is often the case, that the component have fairly good thermal conductivity all over it, and significant thermal mass

Consider some big transformer with steel laminations and copper winding. You overload it by a factor of 2x..5x or so. The copper winding will take a while to heat up. If the overload is moderate (~2x), during this time there is also sufficient time for some of the heat to also transfer to the laminations (which heat generation is only affected by the input voltage, so is not overheating by itself), so slow down the heating even more

Within the time it takes to heat up (and quite uniformly), you have sufficient time to touch its external surface and evaluate its temperature, or a simple bimetallic switch to click, and switch off the power before the temperature exceeds the allowed max for the insulation



With some passive components, damage starts immediately when the abnormal condition is applied, but the component does not fail right away in a way you can see

Consider a garage door opener motor with a start capacitor. The capacitor have a limited life rating, probably in the single digit hours. In normal operation, a second of its life is spent for every start. A capacitor with just 4 hours rated life, will last 10 years opening and closing the door 2 times a day every day

If you power it on with a stuck motor, it won't take long before the capacitor have significantly reduced capacity (and not much more longer before it fails spectacularly)

You might come to the garage after smelling hot windings from the motor (not the capacitor) to find the door got stuck before fully closing, fix the immediate problem, check it once and see that everything is ok. Sometime later, the door will start getting stuck again and again, as the capacitor has now reduced capacity and the motor does not have sufficient starting torque

The capacitor might not even be hot to the touch during the event, as the damage is related primarily to voltage and not temperature



With many electronics, the damage (or start of cumulative damage, as in the case of capacitor) to a specific component may be immediate

Consider a switching power supply, which contains a transformer and a MOSFET

A short circuit (or a very high overload) on its output, may cause the on time to increase above safe limits within a single PWM cycle. This is commonly anything from 1/30000 of a second or less. The thermal capacity of the MOSFET die is tiny - It will vaporize and explode maybe in 1/1000's of a second, before temperature on the PCB near it even starts to rise. It will be impossible to protect it even with a thermal probe directly soldered to its tab

Pulse-by-pulse current limiting protects it in this case because it monitors the actual current and not the temperature. When a short happens, the control IC is fast enough to cut it out within the 1st cycle, right as the current crossed the set limit - Before it actually got to the full short circuit current value

Thermal probe is still an important protection, but for slow things like mild overloads or being used in too high ambient temperature



Consider a MOSFET which is controlling a relay coil. The flyback diode is missing

With each switch off of the relay, a voltage transient is generated, which actually does blow off a small bit of the MOSFET die every time. The MOSFET may be oversized enough to be able to sustain some damage before it fails and stops working, but the damage starts from the 1st time

A flyback diode prevents the generation of the transient in the 1st place, so the same relay with the same MOSFET can last for unlimited time



The IS ballast does not care about the filament being connected end to end - All it needs is for the filament surface to be emissive so it can conduct into the discharge

If the filament surface is not emissive, it will either take high voltage to force current to flow anyway (and by this causing high power dissipation at the filament, which will melt it, and overloading the ballast), or be unable to force current out of the filament, so conducting only in the opposite direction (so rectifying the ballast output to DC, which then leads to the ballast either cutting out, or the available voltage on its output to drop)

In both cases, the ballast may start "flickering" at high frequency and/or hard switching, generating all the effects you noticed

I was recording an EOL tube running on a GE 2x 40w instant start ballast, and the ballast produced a high pitch buzzing. After recording, I saw a lot of artifacting on that video, like blurring and glitching.I also tried using phone cameras to record the EOL tube on that ballast. In my phone the video was having issues and glitching. Even on an old Iphone 6 the video was getting blurry.

But when I recorded a good tube running on the same ballast, none of those cameras had interference. The ballast's buzzing noise was also gone.

Was the buzzing from the ballast affecting my cameras and causing interference?

I think what you see is not the actual electrical interference, but an artifact of intermittent light on camera rolling shutter readout and also on the camera exposure system.

Though, igniting lamps do in fact emit some HF hash. I was impressed when lighting up G12 MH lamp screwed up the transmission on my headphones bluetooth, never thought interference generated goes that high, right into gigahertz range.

The things like MH starting may not interfere with the 2.5GHz directly, but problems with modern consumer electronics uses to be interference with the capacitive touch sensor "buttons"...