Some time ago I visited in work related matters in this military bunker constructed deep underground. I've visited there few times over the years and now I noticed that some of the fluorescent lamps had been replaced with led tubes. The tunnels are lit with 2x36w fixture where another tube is operated by motion detector while other one is usually off but can be switched on by manual switch. Magnetic preheat.

The motion detector side had been lamped with led tubes while the manual side had been left with regular fluorescent. I asked about it from the guy that was keeping an eye on me while I did my work and he said it probably was because led light might be more sensitive for EMP pulse.

That made me think if there are differences how well different light sources can hadle EMP if we assume that the power source is not destroyed by the pulse. Would led lights be more prone to die than conventional lamps with magnetic ballast?

The other reason that popped in my mind was just money saving. Replacing something that was rarely used was unnecessary so they replaced only the ones that gave some real change in electricity consumption.

Some time ago I visited in work related matters in this military bunker constructed deep underground. I've visited there few times over the years and now I noticed that some of the fluorescent lamps had been replaced with led tubes. The tunnels are lit with 2x36w fixture where another tube is operated by motion detector while other one is usually off but can be switched on by manual switch. Magnetic preheat.

The motion detector side had been lamped with led tubes while the manual side had been left with regular fluorescent. I asked about it from the guy that was keeping an eye on me while I did my work and he said it probably was because led light might be more sensitive for EMP pulse.

That made me think if there are differences how well different light sources can hadle EMP if we assume that the power source is not destroyed by the pulse. Would led lights be more prone to die than conventional lamps with magnetic ballast?

The other reason that popped in my mind was just money saving. Replacing something that was rarely used was unnecessary so they replaced only the ones that gave some real change in electricity consumption.

Discharge lamps with magnetic gear are for sure more robust than LED's and electronic gear when jt comes to EMP's

Handling EMP is not about those components alone, but about the way how the wiring is done. The wiring is the antenna picking up the pulse, not the electronic itself. Yes, you have to be able to deal with saturated inductors (due to strong magnetic field) for some time, but I do not know any post 80's SMPS design not able tpo survive it (not all are able to operate during the pulse though, but for most applications that could be tolerated)
With good wiring installation layout the wires pick nothing significant, so even an otherwise sensitive electronic has no issues. With bad layout you have 100's kV everywhere and there is nothing that could survive that.
The thing is, the ratio of voltage robustness between different technologies is about 5 or so (the most resilient vs the most sensitive; so about lets say 12dB), but the differences between a good and bad wiring layout could easily lead to voltage ratios of 1000 or more (so 60dB or more). So the extra 12dB of the "old fluorescents" wont save you if you have 60dB overvoltage.

And by the way the most vulnerable parts of the infrastructure are mainly the magnetic components: They saturate, causing effective short circuits in the system and those blow a lot of things up in the installation. The second most sensitive thing are the high voltages induced on the long stretches of above ground wires. Because for overvoltages the system has a defense (all the spark gaps, arresters and so on), but for core saturation it has none at all. The magnetic components originate from two sources: First the plain magnetic field directly at the place of the transformer, second is the low frequency component induced into the long wiring. The first is virtually impossible to fight (the normal Earth magnetic field is already stretching all the shielding to its limits, but such component would "need" the detonation to be really very close, so there would be other things to worry about way more), the induced low frequency voltage is possible to fight only with rotating the wires along the lines, so it cancels out. Note, for this component even a moderate solar storm and consequent "wave" in the solar wind is known to cause huge problems, when not known in advance (few minutes warning time uses to be sufficient and that is, why we have space devices like SOHO - to give us a warning something is up on the sun; the solar wind travels way slower than light or signal from the space probe, so the difference gives us enough time)

My guess the only reason they kept the old fluorescent was, only the PIR controlled one is used, so consumes energy and maintenance, so they just save the money on replacing the others.
Plus on the PIR one you may expect a lot of ON/OFF cycles, which are not handled that well with fluorescent.
And converting to high efficacy LED (with further energy savings by the PIR) has there quite good reason in case the shelter is really to be used: If you have limited supply and no external power, the energy you have available becomes bloody expensive: The batteries are not unlimited capacity and the generator fuel tank is not unlimited either. Plus in many cases you even can not start the generator (or may use it only in very limited way; because it needs air and generates exhaust), so the only energy you have are the batteries. And then the energy saving light makes the difference to have the light available for just a day or for a week or longer.