HSL across Long Island and then the Sound might be cheaper...
Except the LIRR mainline is very nearly as bad for service density, doesn't have a single
intercity catchment as big as omitted Stamford (remember: Stamford outslugs New Haven by wide margin on Acela-specific boardings, so commuter vs. regional vs. HSR demand
isn't scaling here at the same rates for each service tier). And most of the paths they could trace across the North Shore inlets to a tunneling trajectory ended up being curvy as hell. NEC FUTURE couldn't come up with a cross-Sound routing that kept its service margins, so that idea is pretty much dead.
As an idea that's always going to tickle the imagination just like cross-Sound expressway bridges/tunnels have tickled the imagination for 70 years it will continue to be a perennial conversation piece. But it never ends up adding up to sum of parts on closer analysis. I don't think we're done studying this at the big-picture level per se...but each additional look doesn't end up pushing it in a direction that starts producing "Eureka!" answers for the toughest questions of un-served demand that come up each time. Much like three-quarters century of cross-Sound highway bridge vaporware, a conceptually always-compelling idea that nonetheless becomes self-inhibiting to planners as soon as the briar patch of fine-print questions prove to have no easy answers.
Usually additional tracks on the Metro-North are discussed in terms of a 5th passing track, and elsewhere in Connecticut, the occasional two track bypass (Groton-New London)
Grade separating "SHELL" interlocking is the big one. Right now the Hell Gate/Penn side used by Amtrak and to-be used by the coming minority of MNRR Penn Station Access trains has a painful 30 MPH restriction at the flat junction, while the GCT side used by the MNRR New Haven expresses has a 60 MPH restriction. Crossover layout is squished way back of the junction making Amtrak have to do the hardest crossing work, with high chance of getting stuck behind a slowpoke Stamford-or-New Caanan local (the biggest by-far share of overall west-of-Stamford service) en route to New Rochelle Station because they go over the junction 30 MPH slower. New Rochelle then has a funky platform layout set by those constrained Penn vs. GCT crossovers that explains in part why Amtrak stops there for their lone before-Stamford intermediate but the MNRR New Haven expresses have to keep skipping on to Greenwich. There's zero flexibility, and bad timing at the junction causes bunching galore that has little/no room to reset itself before Stamford. This is the largest-share source of "bad behavior" problems where express trains making only 1 intermediate stop before Stamford should inherently be able to keep out of each other's way...but often don't and get entrapped in bunching hell.
"SHELL" separation would clean that mess up thoroughly by trenching the Hell Gate/Penn tracks underneath to direct-feed into the center express tracks, while moving at least 40 MPH faster than today. That way AMTK never has to worry again about getting stuck behind a slowpoke all-stops local to New Rochelle and having to make corrective action before Stamford. Acelas would also gain enough extra speed to beat the New Haven express into the junction so pure chance doesn't put it behind a Greenwich-stopping express while it runs nonstop to Stamford; dispatch can time the Acela for priority over the commuter express without penalizing the express's schedule because a few MPH throttle is all it takes.
Northeast Regionals would then be able to run up ever closer to the taillights of a NHV express at the junction since they'll be stopping at New Rochelle while the NHV express opens up a spacing lead into its Greenwich stop that is fail-safe from being squandered before Stamford. And there are supposed to be TBD station layout changes to New Rochelle Station for more platform tracks to assist with the flexibility...probably changing the side + island layout to touch all tracks instead of 3 out of 4 like today.
In ancient times SHELL to Mamaroneck Station was ex- 5-track territory when the long-defunct New York, Westchester, & Boston line co-mingled with the NEC before turning north to White Plains. No appreciable stretch of 5th running track is going to be possible with the space required to grade separate SHELL and redo the New Rochelle platforms. North there's an extremely rarely used layover yard, which may be reanimated for a small slice of Penn Station Access 'city zone' trains that surge frequencies at the new Bronx intermediates to quasi- rapid transit service levels but terminate inside NY out of the way of all other trains rather than continue to Stamford. Those should, because of the station reconfig and slack space just north, be able to slide in and out with no conflicts to any other traffic. Track 5 space is then fungible for a very short distance to help with extra sorting space coming in/out of New Rochelle for additional resiliency.
Other than that you simply won't find the room for extra tracks. I-95 before the CT state line cannibalized the NYW&B's extra side-running tracks on the ROW when it was built, which is why they are bolted together here. You might also have a short sorting length you can squeeze in at Greenwich Station for additional resiliency on the express single-file where NHV expresses are stopping but all of Amtrak is non-stop. That helps a little, but the lengths of space available are so uniformly small that they slot in the category of micro-adjustments...nothing remotely game-changing. The whole way to Stamford the NEC is either sharing cut/embankment with I-95 or they're splitting for insignificant lengths so 95 can bypass a downtown that is wall-to-wall massed against the NEC. On that whole stretch the most consequential dispatch assist is replacement of ancient Cos Cob drawbridge with a faster-moving, adjustable-height, more reliable lift span that slashes bridge openings sharply back on the stopwatch and also sports a thinner/sturdier deck than the current bascule to let smaller speedboats slip under without triggering an opening. Of the 5 movable bridges in SW Connecticut, Cos Cob is responsible for 50% of all New Haven Line maritime openings because of the heavy pleasure boat traffic in Cos Cob Harbor. Norwalk--fully-funded for shovel-ready replacement as a new lift--does 25% (mostly cement barge traffic), and the other three split the remaining inconsequential 25% on the lightest-traffic Bridgeport-New Haven segment. A new Cos Cob lift would have somewhat lower total quantity of openings with the extra couple feet of speedboat clearance, and allow dispatch to pack trains much tighter around openings because lifts are inherently fast and there's rare enough tall-mast traffic here that it would rarely need to lift all the way. As a "good behavior" and dispatching precision aid that state-of-repair replacement ends up being more meaningful than any short Track 5 sidings you can cram in there.
So in the end, doing the "SHELL" grade separation megaproject and a whole lot of unsexy stuff like bridge renewal hits the target where NY-Stamford can live just spiffy inside a 'Superduper 2040 HSR' service universe. Very little of the bucket list here requires the level of overthinking that the Shoreline bypass or 200 km/h NY-DC does...which is perhaps why there's such temptation to overthink MNRR territory regardless. Really, it 'works' in context of all of the service goals of the NEC at-large if trains are simply better-behaved on the express tracks through the SW CT megalopolis. Those 4 intercity stations have to be served in service-fileted fashion in order to handle the loads or give anyone in that part of the megalopolis fighting chance to survive the last-mile trip to the station. So the goal was always "marathon not sprint": keeping even pace where all demand actually stacks rather than loading up for some singular NY-NHV sprint that leaves an outright majority of the megalopolis' demand laying on the table.