IMHO the sound should be like a real crack of thunder, the one crack, and that distance from that original instantaneous crack should reverberate. Like in the real world.
The real world has a way of being really complicated. I happened to come across a book describing the physics of thunderstorms and lightning recently, and the amount of stuff I did not know about the topic before was embarassing.
Reality starts with a (roughly cylindrical) shockwave emerging from the discharge channel. This is not a sound wave initially because it goes faster than speed of sound, but after a few meters it attenuates sufficiently to become one.
Even standing twenty meters from a lightning strike, you would hear rolling thunder, because the sound from the lower end of the channel reaches you basically now, whereas the upper end of the channel at the cloud base is perhaps some 500 meters away from you, and you would hear that sound with a delay (of course, it would be lower in volume, as it would have attenuated, and you'd be blinded and near-deaf anyway....)
Standing further away, the horizontal distance eventually becomes as much important as the vertical distance - this part is simple geometry.
Standing yet further away, it matters that sound propagates through the atmosphere differently dependent on density, in particular sound waves are systematically deflected away from the surface. This is the reason you can't hear lightning for more than 30 km distance - it doesn't attenuate that much, you could theoretically hear it, but the sound waves can't reach you - if you were in the air, you would hear the sound farther.
A lightning strike also isn't a simple event - there's the fanning out of forked channels from the cloud, then the upward strike of the main lightning bolt to meet them, a downward-racing re-ionization of the channel, another upward main bolt, can be a sequence of up to ten events - which also draws out the sound.
So what you hear is a function of geometry, distance and altitude not just in volume, but qualitatively, and you would actually need a few different sounds to do it justice.
Please note that the sounds in the atmosphere reverberate. Off of mountains, clouds, buildings, road surfaces.
I think we may leave clouds largely out of this list... But it's really not the main reason thunder rolls and is not a sharp crack - that is just different distance to the various elements of the discharge channel, leading to staggered arrivals of the sound waves.
It might be possible to handle say like a global thundersound much like we hear AI/MP sounds. Exterior to the actual aircraft code.
Some people (Erik included) seem to favour this solution, I frankly do not - what you can hear inside a cockpit depends on how sound-proof the cockpit is and what the ambient sounds are. In a single-prop GA plane, the noise level in the cockpit is extremely high when the engine is running (last time I did this, we had headsets on and could talk only via mike) - I doubt you'd hear any rumbling of distant thunder. Hanging beneath a hangglider, I you would on the other hand. The cockpit of the Space Shuttle is sound-proofed to the extend of damping the inferno of multiple rocket engines 30 meters away (which are still rather noisy from a mile away...) - I'm fairly sure if you're inside with the doors closed, you can't hear a thing of a thunderstorm.
So while it may be possible, I personally don't think it would be desirable. My opinion is that we should store the environment ambience sound files in a common location, but that they should be used aircraft-side because the aircraft developer knows best what can be heard in-cockpit.