The best you can do for the New Haven Line is getting it up to full state-of-repair. Fix the bridges, knock out a couple deferred maint speed restrictions, increase the reliability of the signaling and electrical, etc. Because the signal density is what it is and Metro North mixes so many expresses with locals that there's never clear passing tracks to rev up AMTK speed except on the far off-peak, state-of-repair is the best you can do. But it matters...a lot. Raw speed isn't nearly as much a quality/quantity of service limiter as on-time performance. The more reliable an OTP they can achieve, the more trains they get to run and the less they have to over-pad the schedules for when they clear New Haven, which will bleed minutes off the schedules.
The actual-factual speed increases come from elsewhere. Built-up downtowns with dense commuter rail stop spacing are always going to be slower, curvier, and more congested than the swamps of New Jersey. You can't avoid that because that's where all the people are. NEC FUTURE needs to be beaten with a shovel to get that in their thick skulls that bypasses of population centers defeat the purpose of the service they hope to achieve. Increasing speeds to shorten schedules isn't something to hyper-focus on with the New Haven Line. Schedule improvement gets achieved by running up the score through the Jersey and Delaware swamps by opening up more 165 MPH territory, finding a solve for Metuchen curve, and so on. And by putting money where it's needed into zapping the Wallingford and Meriden grade crossing clusters + associated speed restrictions on the Springfield Line, and by speeding up MBTA territory where a couple decently large-mileage stretches of the Worcester Line can sustain 90 MPH while leapfrogging commuter rail locals.
The New Haven Line is what it is: an unbroken megalopolis where all the people are. Your goal is to get it in tip-top reliability. Do that and it gives plenty of service enhancement in the form of vastly better OTP. The raw performance improvements come from building up the head of steam in less populated/more easily straightened areas to build up that time savings cushion for the slower New Haven Line trip. They work hand-in-hand...you rack up savings in the Jersey swamps, but you sustain them by having full state of repair and dead-on accurate OTP in Connecticut. It's more nuanced than just a one-dimensional focus on "We have to have 125 MPH or better everywhere or we've failed at building first-world HSR." That kind of simplistic over-focus is what gets garbage like NEC FUTURE proposing to bypass the shoreline cities where all the people are to blow $100B TBM'ing dozens of miles of tunnels through the rural boondocks. places few people live. Running up the score where the score can be run up, then plugging leaks in the max-density jungle of the Route 1 megalopolis so there's as little attrition as possible when coasting through where all the people are is how the 'Dream NEC' gets done right and how the very divergent traffic profiles of these differing NEC pieces end up fitting together fluidly and functionally.