When I read the news of the exploded Dell batteries, my first thought was a problem with the wiring/welding of the cells. It is a fullly automated process and it undergoes very little QC, so there is room for large scale mistakes.
As Tom pointed out, my hypothesis is not the only possible cause of the problem. Let's see.
A laptop battery is made of a series-parallel of Li-Ions cells. Every group of paralleled cells is monitored separately during charge and discharge, and the individual capacity (of each group) is measured at each cycle. Any variation of capacity is written in a flash memory, together with the number of cycles and other information such as voltage, temperature etc, as specified in the SMB-forum protocol.
The laptop communicates with the battery with a serial syncronous connection. Since every battery contains one or two NTC sensors plus a thermal fuse, every problem is usually communicated to the OS, which shuts down the computer.
As long as the electronic controller is working, cells cannot become unbalanced both during charge and discharge. The effective capacity of a laptop battery is directly related to the capacity of its weakest cells.
Now, the firmware can lockup, like any other code-driven interface. In this case, the serial link with the computer is interrupted, and charge/discharge is no longer possible.
The MOSFET could short, and in this case an error flag would be raised and the battery deactivated. Same for overtemperature, or a sudden change of the charge/discharge curve of the cells.
Mac computers allow the reset of the battery from the console. Win computers require a specialized software to reset the battery.
So, in principle, I would exclude an electronic problem, since there are too many controls in place.
Of course, this is only a line of thinking.
A defective cell could be a problem. Li-Ion have their mechanism of failure. Precipitation of the cobalt, as Tom said, is one. An excessive precipitate, due to an excessive de-crystallization of the Lithium Cobalt Oxyde, could lead to a short-circuit of the ionic separator in the cell. Each cell is directly paralleled with one or more other cells, with no supervision at all from the uP.
This would make the other good cells set the defective cell in flames with no possibility of recovery.
The fact is that the precipitation occurs after the 500th cycle, and the battery is programmed to stop working after 300 full cycles.
Of course, this is valid for good cells... in the case of the exploded Dell batteries some cells may have born defective.
Please take a look here:
http://www.sbs-forum.org/specs/sbc110.pdf
Anthony