The maximum length of holding the train really should depend on what the "pad" was between scheduled arrival at the other end and the scheduled departure of the connecting train(s) there, as well as the average or expected delays anticipated on the trip TO that connection point. If there was a 5 hours time pad between normal arrival at the connection point and departure of the connecting train(s), and the OTP of the train was realistically expected to be good, then they could reasonably wait more than an hour. But all those factors should be weighed carefully before deciding. Obviously, whatever wait time they added would inconvenience passengers who had destinations before the end point of that train, but as long as the connection was made AT the end point, the wait would not inconvenience the connecting passengers at the end point. If the connecting train at the far end was running several hours late, that would allow waiting even a longer time. However, if we use the specific example of a NOL-JAX train waiting for the late arrival of Sunset eastbound into NOL, any late departure of the train heading for JAX would inconvenience passengers into Mobile, Pensacola, Tallahassee, etc., even if the passengers connecting to a Silver Service train successfully made their connections. All of these sorts of decisions involve trade-offs. SOMEBODY is always going to be unhappy with the result.