The afternoon train covered the 340 mile run in 3 hr. 59 min but the overnight run took 8 1/2 hours.
Well, there's your problem! I don't think an overnight train can possibly be successful when you can make the trip by day train in 4 hours. If driving or taking a day train already takes about 8 hours, then they start being worth considering. A fast day train is always preferable. (I actually don't see many good overnight routes in Canada. Windsor-Montreal and Toronto-Quebec City are the right length, but I suspect the nature of the travel market means these are low-demand city pairs. Montreal-Moncton should be good; Montreal-Halifax is too slow.)
There are a bunch of good routes in the US, though. Driving from Chicago to Minneapolis is 6.5 hours -- if you don't stop and don't hit traffic. So really, 8 hours.
The airplane? Well, my estimate says MSP-CHI takes 4 hours including all the trouble to get to downtown Chicago, security, etc. (I tend to be a pessimist on flight times. The nominal flight time from Philadelphia to Ithaca is 1 hr. 5 min. It is frequently faster to drive, at 4 hours, thanks to security, waiting for the plane, delayed planes, etc. These short puddlejumper flights are all "overhead time" and are not worth it.) So sure, a bunch of people will choose the flight -- but a bunch of people will not.
For Minneapolis-Chicago, next-day tickets are cheaper on Amtrak than in the air. The *roomette* is actually cheaper than most of the airline tickets. The roomette price could probably be brought up quite a lot and remain competitive -- all that is needed is that the train run on time.
(3) It's quite possible that we end up with a long-term scenario where the business community ends up on high-speed trains that tend to operate with fat margins and fares most of us can't afford (i.e. the Acela). Everyone else ends up on slower Regional-type trains, and overnight runs exist to either (A) connect corridors or (B) serve longer runs that the high-speed services would take a decent portion of a day to cover (mainly operating over routes that would take 8-16 hours to cover with conventional equipment, and which high-speed trains would still take 5+ hours to cover...basically picking up a mix of traffic that can't do an all-day train plus odd-and-end traffic that either needs to leave a given place pretty late or arrive pretty early compared with the other HSR services).
This is pretty much what I've been thinking. The overnight trains will end up on routes which are just a bit too long for good day trains, but where airplane service is unreliable, infrequent, expensive, and is taking up half the day anyway. There's actually a number of these potential routes out there; the best are through "second tier" cities like Syracuse & Rochester, which the airlines dislike serving. And the airplane service in such places isn't going to get better; that time has passed.
It's important to realize that these "best overnight routes" generally overlap the good corridor routes, for demographic reasons: they run through the same "second tier cities". For shorter trip times, you run day trains (example: Buffalo to New York City) and then once the chain of cities is getting a bit too long (example: Buffalo to Chicago) you *also* run an overnight train.
By all means, make improvements on the "corridors" your top priority. It will only benefit the overnight trains. If anyone remembers their history, this is how the passenger rail network was built up in the first place: little corridors were chained together, and eventually overnight trains were overlaid onto the longer of the routes. It would have been considered bizarre to run an overnight train without first running a bunch of local trains.
The formation of Amtrak was done in an ass-backwards manner, and the lack of underlying and connecting shorter routes hurts all the long routes. The shorter routes have since been rebuilt in a number of places, and it should make a difference.
I'll be particularly curious to see what happens in Denver, where the first urban rail line got to Union Station only in 2002, but in 2018 there will be seven plus a bunch of intercity buses. Denver's been bleeding Amtrak ridership for two decades, though I can't find the year-by-year numbers I'd like to look at. A lot of that is due to the unreliability of the Zephyr, of course. (I say, break the train at Denver to improve the OTP east of Denver.) Part of it is due to the cancellation of the Pioneer in '97, part of it is undoubtedly due to the loss of mindshare when Ski Train went away in '09, part of it is due to the "temporary station" since 2011. I do wonder what will happen when there are huge numbers of commuters walking through the station: I expect a fair percentage will say "We can get to Chicago by train? We should try that." Of course, if the train can't run on time, they won't try it twice.