LA's Rotems were designed specifically for crash energy management in a push-pull system. They were pioneers because they had to be after several bad wrecks that might have been more survivable with CEM cars - though head-on with another train at 50+ mph is probably not survivable for those up front anyway. Metrolink's problems were bigger than just ordinary grade crossing wrecks.

EMUs and DMUs designed for use without restrictions are considered locomotives by FRA and have to meet the same crash standards. CEM helps to handle that without grossly increasing the train weight.

Grade crossing wrecks even with cab cars aren't usually a huge problem for the train, unless somehow a high-speed derailment occurs. CEM can help, but US CEM is designed more for train-train collisions than for grade crossing hits. Wrecks like Amtrak in Nevada, where a vehicle runs into the side of the train, are basically unmanageable - you just have to hope that the hit isn't in a place that will cause derailment or start a fire in the train; and whether or not the train is EMU doesn't really matter.

Volpe (US DOT's research center) did a bunch of testing in the early-mid 2000s that led to the Rotem cars and use of CEM. See for instance:

[onlinepubs.trb.org]
[ntl.bts.gov]

And for locomotives too:

[ntl.bts.gov]

And of course the obligatory Youtube video of the crash tests:

[www.youtube.com]

Cheers.