> But don't we have a window of minutes to shut down the system in case of an earthquake?
No.
> I thought we could get advance warning from P waves. (This doesn't work if you're in the epicenter, unfortunately.)
Yes, but the advance warning is on the order of seconds, not minutes (per Wikipedia, for deep, distant, large earthquakes, 60-90 seconds is possible, but that's still at most a minute and a half.)
Even so, I thought that the failure mode for tube misalignment of hundredths or even tenths of an inch would result in stoppage, and perhaps damaged "track" and cars, with trapped passengers. This isn't so different from the performance of high speed rail during natural disasters.
No.
> I thought we could get advance warning from P waves. (This doesn't work if you're in the epicenter, unfortunately.)
Yes, but the advance warning is on the order of seconds, not minutes (per Wikipedia, for deep, distant, large earthquakes, 60-90 seconds is possible, but that's still at most a minute and a half.)