And that is the difference. It is one thing when you have a lot of equipment and options, such as frequency and support. When you have a sparse network, with trains isolated and at the behest of freight operators and outsourced assistance, uniformity is a good thing.
When there is a problem with an Acela due to a problem with a coach, the whole set is shopped. Setting out a car should be a quick and easy process. Being able to add equipment (if necessary) should be a quick and easy process, particularly when you don't have an extensive network.
I hope it is.
Actually, it is the Acela IIs which will be truly articulated like the TGVs using Jacobs Trucks.
I think you are quite correct in your observation that Amtrak operates on thin margins of equipment, unlike the much much larger passenger systems elsewhere. In a manner of speaking, they have a much deeper equipment bench for all sorts of equipment than Amtrak does.
There are two more observations I'd make based on what I observe in Europe and India.
1. Even when articulated sets are used the drawbars connecting the individual cars within a set together are relatively easy to couple and decouple, so changing out cars within a consist often is not much harder than if they had standard couplers, and are similar in complexity to chain and screw couplers which is still not that uncommon in the rest of the world anyway. Of course, the more permanent and hence harder to separate the coupling is the more seamless is the customer experience of passing from one car to another. It is a tradeoff, and a decision to be made based on what factors are more important to the operator.
2. Typically EMU and DMU sets are three or four cars long, and trains are built stringing together two to five such sets. So even if cars cannot be replaced, all that happens is things are taken out of service in units of say, 225 seats instead of 75 seats, and that together with availability of spares in units of 225 seats instead of 75 as standard units for example, works out just fine. And the 3 or 4 car units then can have full width vestibules giving a sense of airy spaciousness to the customer, with only inter-unit vestibules being narrower.
In India so far even when cars are connected by drawbars within a set, no advantage is taken to provide a more seamless experience to the customer. That is changing with the recent T-18 distributed power prototype, which supposedly will be deployed on Premier trains, both regional and overnight. The claim is doing so will allow lopping off full two hours from the current 16 hour schedule between Mumbai and New Delhi or Kolkata and New Delhi.
As far as I can see, at least in India, the argument for going with EMU/DMU for regional service is the significantly superior timetable performance of the DMU/EMUs over loco hauled trains. In general they report immediate 10% to 20% overall running time reduction for services that have frequent stops. That really is a big deal if achieved without requiring major track/signal overhaul. This is the publicly stated primary reason for converting regional passenger and express trains to DMU/EMU, and it clearly shows in time tables and customer experience.
I saw a living illustration of it the other day when the train I was on, an extremely capable loco hauled train of 10 cars with 5+kHP Co-Co TRAXX electric engine, was routinely beaten handily in acceleration out of a station stop by an EMU. Of course eventually we caught up with and overtook the EMU since we did not have to stop everywhere, which it did (and our APS was 30kph faster than the EMU's), but the illustration was stark. And it is not like these EMU sets are drastically overpowered. It is just that the power is distributed in 1 in 4 (16 axles in 64 in a 4 unit 16 car train) instead of 6 axles in 46 to 66 (10-15 car trains). There are no trains in India that operate with more than one electric engine (except the few mountain segments where bankers are used just for the climb), or for that matter more than one EMD dual cab ~5kHP Co-Co AC drive diesel engine. There are some that operate with twin single cab ALCO derivative 3kHP DC drive engines.
But bottom line is, the operating conditions and imperatives are different in different environments. US passenger service except in the NEC is very different from the more intensive passenger ops elsewhere in the world.
On the NEC 3 or 4 car sets with one power car coupled together using AAR-H couplers as proposed by NJT is likely to address the distributed power issue without compromising on replaceability of individual cars, and indeed that is what Amtrak might go for. If experience elsewhere holds true in American, which is always a big if, then this could improve timetable performance considerably without massive investment in tracks and signals. So we shall see.
As far as I can see, the two features to go for primarily are:
1. Distributed power - spread the power transmission among a larger proportion of axles to get significantly better performance.
2. Operating Cabs at both ends - to eliminate the need for turning trains on Wyes or Balloon Tracks or unhooking and hooking locomotives at the end of each run.
The rest is just arguing about who calls which configuration by what name.