Extended sound of thunder

I have always assumed it could be put down to a small set of causes:

• The stroke may have a physical extent and geometry that leads to a perceptible duration between the arrival of the sound due to the near part of the strike and that due to the far part. That is, what Georg said.
• If the geometry of the terrain is right you may be hearing echos off of cliff faces, hill-sides, buildings, etc. These obviously come by a longer path and therefore arrive later.
• An active storm has masses of air at different temperatures and humidities. The speed of sound will differ slightly between these regions, leading to refraction at the boundaries. As such there may be lensing going on in the air and again you hear the same sound arriving by multiple different paths at different times.

Add to that the fact that most strokes are in fact multiple discharges, and you can hear sounds that last much longer than the observed light.

I have not gone to the trouble of taking precise data, but my impression is that the rumble tends to be longer for strikes from further away, suggesting that the last two points are the dominate effect.

Other things to look for include

• Frequency dispersion. But if this was important you would expect the pitch of the sound to have a characteristic dependence on time, which I have never noticed.

This is because the different parts of a stroke are be very different in distance.

This

Reverberations and echoes of sound last longer than those of light.

is obviously wrong. Which walls or mountains are the reflectors of those echos? Which rooms do the reverberations?

The 3D shape of the lightning channel over many kilometers of extent can be teased out of three simultaneous sound recordings made from three different spots. The coding is similar to that which produces CAT scans, very abstruse, lots of matrix algebra. Similar systems are used to localize gun shots in urban areas using microphone networks.