Read through the NewBie thread... With two cell lights/simulations, he took one partial discharged cell and one full cell (primaries in this case), and ran them under load until "something" happened...
What was interesting, in at least one of the cases, it was the 100% new cell that vented--not the partially used cell...
Why--possibly the partial cell got hot, headed the "new" cell, and when the cell(s) failed internally, only the "new cell" had enough energy to overheat and vent. So, in this case, it would be assumed that it was not the polarity reversal of the "used cell" that "failed"--but it may have contributed to the venting of the other cell because of overheating.
In any case, the failure modes and mechanisms are numerous and varied. Some are probably happening in flashlights that have not been used for a while, others in flashlights the are even perfectly functioning until they got hot, made noise, "went dark" a few moment before, etc...
There appears to be no one failure mode--But the majority failures in primary lithium cells appear to be broken into to classes of causes--1) Heat--whether because of high current, "weak" cell in string, environmental; and 2) mechanical damage--dented cells from over-tightening, dropping, etc.--once "damaged" the massive amount of current available inside the battery causes internal shorts and I^2*R heating and/or chemical heating.
I have not read the whole thread yet--but most of the failures seem to involve two or more cells in series. Yet, there was, apparently, at least one failure noted (Air Canada incident?) with a single cell light that may have been dropped on the floor of the aircraft by a passenger.
In the end, these batteries have very reactive components and are able to release large amounts of energy in a short amount of time... Perhaps Lithium flashlights need a frangable spot or other venting system to help reduce propelling parts of the lights ("explosions") when failures occur.
Problem is that this will add costs, size, difficulty to change cells, and could reduce water resistance and such...
Problem is real, appears to be rare (unless abused on purpose), and is the cure worst than the problem.
In the end, the current rules that keep the size of the batteries small for use on airplanes and excluding from nuclear power plants may be enough.
In the case where a small "venting with flame and flying debris" incident is not acceptable, either the light(s) need to be redesigned or accept using alternative energy sources (alkaline or other chemistries/constructions).
The neat thing is with the high efficiency LEDs, these alternative sources of power can give just about equal performance with far less risk of "dangerous" failures.
-Bill