At least three times, I’ve seen the signal crossing just east of the Sunnyvale Caltrain station exhibit the following behavior: as a train approaches the station from the west, the signal starts to blink, and the crossing arms drop. The train stops at the station (still not having reached the crossing), to let passengers get on and off. Just before the train leaves the station, the crossing “times out” (the train never having reached it) and turns off, letting cars cross the tracks for about five seconds before the train pulls out of the station and activates the crossing again.
I can understand the pessimal behavior on the part of the crossing equipment; it doesn’t know whether or not the train will actually stop at the station. The crossing is only a few hundred feet east of the station, and if an express train is approaching at 79 mph, it does need to start its signal well before the train would reach the station. Otherwise, there would not be enough time to safely enable the crossing arms. But it seems almost as unsafe for the arms to come up for just a few seconds, without a train ever having passed. I could easily see an auto driver, after seeing the signal come on for a few minutes, then stop, and then immediately start again, without seeing a train, deciding the signal was just broken, and trying to cross anyway.
It seems like the easy solution would be to increase the length of time the signal waits for a train to pass to encompass the extra minute or two this scenario encompasses. Of course, that does add an extra minute or two to the case where the signal really is triggered unintentionally, by a train that does not plan to enter the crossing anytime soon, or when the signal is accidentally tripped.