Along with Bob's litany, note that the Glendale and Oxnard accidents were not actually at grade crossings. In both cases, a vehicle was wedged in place on the tracks away from a crossing (though close in Oxnard, and allegedly not deliberately), derailing the train. Glendale then evolved into train-on-train as the derailment progressed. The crush performance of the Bombardier cars was poor enough in Glendale that CEM research was stepped up and the Rotem cars designed. Now we're hearing that a possibly mis-handled detail makes the Rotems vulnerable to derailment in more common (than train-on-train) situations. That, of course, needs to be fixed if possible. I don't see this as purely (or even mostly) a PR thing.

Actual grade crossing wrecks are pretty common on commuter lines - sometimes several times a week for ML and Caltrain (California drivers?) - but seem only rarely to cause derailments regardless of push or pull operation. Even in push mode, the train is so much bigger and heavier than the road vehicle that it's not a contest. Also, while ML and Caltrain do often contribute to grade separation projects (more so than most railroads), it's the road owner's responsibility for the most part to make improvements to grade crossings or to separate them - and eliminating the grade crossing is the only really effective way to eliminate grade crossing accidents.