This is because when a tant cap fails it fails short circuit and that is a fire risk.All caps can fail short. In fact the most common failure for a ceramic is flex crack which leads to direct short. (My college calls them LECs: light-emitting caps.) That's why ceramics come in "Open-Mode" configurations and Flexible-type terminations. It is one reason why aluminum electrolytics have safety vents (or stress points designed to blow.) So tantalum isn't the only cap that can fail short and it isn't the only one that can ignite.
It is true that a MnO2-based Tantalum can flash. However, when properly de-rated (at least 50% of the largest voltage transient) the chance of failure is almost zero. The failure occurs because of weaknesses in the dielectric which are created when the parts are subjected to reflow oven temperatures.
Polymer-based Tantalums have a benign (no ignition) failure mode as their cathode contains no oxygen. Of course their low-ESR probably makes them a bad choice for this application.
If you ever want to get a circuit UL approved they will not allow the use of this sort of capacitor in this situation.I have never heard this to be the case. Proper testing after the board has reflowed can eliminate any safety hazard.
Tantalums don't fail over time. They only fail the first time the dielectric is subjected to a voltage it cannot withstand. If after reflow the maximum voltage the cap will see is applied, and it does not fail, it is extremely unlikely to fail during its life: the same as any other component.
Note: I work for a capacitor company.