This article I found about magnetic vs. electronic ballasts makes so many incorrect statements that it hurts my feelings. I won't even get into it here; I'll just let you read it for yourself.

http://www.firerescue1.com/fire-products/emergency-lighting/articles/291329

Well, the matter is there really extremely simplified.
I guess you are judging the safety by seeing the components in an electronic ballast really melted down or pitch black, clearly showing very high temperatures there.
But have you really seen any single fire really going out of such failing electronic ballast (don't mix with CFL's)?

But there is a catch: Although the electronic failure involves way higher temperature of the failing components (therefore it looks so scary after disassembling), these components are very lightweight. So even when incandescent hot, the heat energy in them is very limited. So just the sheet metal of the fixture body is way sufficient to reduce the temperature to low, safe level, so nothing could catch fire from that anymore, all the heat is well contained within the fixture (and in all cases I have seen even here within the ballast case; that does not mean that case is intact).
On the other hand if the magnetic ballast is failing and overheats, you have few pounds of overheated material. That mass means that heat goes quite easily behind the fixture metal work, so reaches the structure behind. And if that is combustible, you have quite serious fire problem. And here on LG were already presented few cases of quite severely overheated magnetic ballasts, not that far from really igniting something. And these were mainly the European style series chokes, so the simplest ballasts.
To prevent such fires, indoor ballasts are required to be protected against such overheating. But if that is some kind of switch, that switch may fail. And most likely it will fail, when it starts to operate. And that will happen, when the ballast starts to overheat. So even when these cut outs did prevented many fires, still quite high risk remains due to their failures.
The lightweight electronic ballasts just do not have enough mass to store so much heat before at least the circuit fusing acts (as really a last resort)
So in the fire safety aspect, the article really is correct.

Of course, all that assume normal indoor environment, mainly no flammable vapor or gases in the air,...

Presently was able to get a 250W mercury Westinghouse Silverliner lamp fixture from our yard light redo project at work.Its ballast must be orig-light still works just fine!!!I wonder if a modern solid state electronic or digital ballast will last over 50yrs???A companion fixture at 400W has a blown ballast-will see if I can find a Mercury/MH 400 W ballast I can use in it.Just have to find the ballast in my "junk room" and put it in the light.That way I could use mercury or MH bulbs in the fixture.

@tolivac: The article is mainly (and my remark above ONLY) about safety, not that much (the article)/not at all (my response) about reliability.

Of course, if  the electronic is made so, it is by price comparable with the magnetic, it will fail way sooner, or better to say way more frequently.

But the thing is, once the magnetic fails, it is way more potent to cause some serious fire, just because of the heated mass.

when a component fails catastrophically-don't really think it matters if its SS or magnetic-arcing or sparks thrown from the component will most likely cause a fire.Then it really doesn't matter what the component is or made from.Have seen some arced SS devices that could certainly cause a fire if the sparks and hot material from it landed into something flammable.If the SS ballast or magnetic ballast is inside and enclosed completely in a fixture-don't think it will be a problem.And--Most commercial/professional fixtures are individually fused so the bad light is removed from the circuit without having to rely on a circuit branch breaker or fuse to interrupt the circuit-thus plunging the area lighted into darkness.The fixture fuse will blow removing that failed light from the circuit without affecting others.Beleive its a code requirement under some circumstances.Many times SS devices fail by shorting-this can cause related components to burn and arc,catch fire.Glad failures of this nature in SS or magnetic devices are VERY rare.

The presence of the danger indeed depend on the exact failure mode.
If a failure result into a hard short circuit of the mains, it may do quite loud "bang" and trip the breaker, but if the thing is designed reasonably (so it contains the eventual sparks or so; that is not that difficult, nor expensive at all, just a cover on the ballast and mains connectors does the job well), it is all what could happen.

The main problems are those failures, yielding excessive temperature on surfaces, where these high temperatures were not expected, so could touch some easy to ignite material (e.g. fixture surface touching the wooden ceiling). With the hard failures the dissipated energy is very little, so does not heat up the fixture surface at all. So such hard failures are not of the main safety concern, they are handled pretty well by the regular circuit fusing.
The same is usually the result of an electronic failure: It's components are small, so either they dissipate only very limited power (so no overheating of the fixture surface), or they (alone) immediately overheat so severely they hard fail (so the "bang" and fuse/breaker tripping) way before they could heat up the ballast cover, so again no fixture surface heating, so no big deal. Of course, most of the the ballast internals are toasted, but yet the cover stays intact, so no threat to the surroundings.

But if you have e.g. a starter short circuit and consequently ballast failing due to overheating, what you get is not a hard short circuit, but a kind of "soft" resistive failure. Because it is resistive, the current is limited, so the fusing does not respond (the fixture fuse respond to twice the nominal current and are usually specified with nominal current of twice the lamp current), so at least until about 4x the lamp current (assume single series reactor) the current will continue to flow without any restriction. With a 36W fixture that means power dissipation easily 400W before the most sensitive fuse could respond, with 2x36W we are talking about 800W on a single ballast choke. That is by far enough to heat up the ballast coil so, the heat reaches the wood under the fixture and it then catch fire from the hot fixture surface.
It is the inability to distinguish the soft, but yet dangerously dissipating failures by the overcurrent fusing, combined with the inability of the fixture cover to isolate that heat from it's surface, what makes the magnetic more dangerous.

In fact at least here in Switzerland all the classic magnetic fluorescent fixtures without a back metal cover between the ballast and the mounting face have a warning statement to not mount the fixture on flammable materials.

By the way on the magnetic ballast normally there is the delta T on normal condition (usually 30-60°C) and under abnormal condition (=stuck starter) and it's not rare find in this case delta T of 150°C or even more! The life of the coil at such temperatures is quite short and when the isolation start to fail the temperature rise to hell...

But the problem is, way too frequently these warnings are ignored, mainly with an argument like "It is just barely warm, even after tens of hours continuous operation, so it is nonsense".
Well, then one time the ballast fails...