I mean, Delta had 133 routes in the latest DOT report (June) of flights that operated at least 5X a week that were late (by A14 standard) at least 50% of the time in June. Obviously those routes change month to month and so I'm not saying Delta has routes that are perennially under 50% on-time performance, but I can guarantee you they have routes where they hare sustained multi-month periods of <50% A0 performance - and if you define a misconnect as "arriving at the gate later than T-15" (since you have repeatedly said closing the door at T-12 to T-15 is not closing early) then those routes would absolutely have a >50% misconnect rate when booking a connecting flight at MCT.

I'm not sure why you think a 50% misconnect rate is absurd on some routes when booking MCT and defining a misconnect as arriving at the gate after T-15. Even when the plane gets parked at A0, that leaves only 20 minutes to get the boarding door open, get off the plane, and get through the airport and to the gate. Using Atlanta as an example, there are plenty of routes that Delta has A0 rates around 50% (i.e. median flight arrives exactly on time), and any flight that goes mainline -> regional will generally mean a transit from T/A/B to D. From T/A that is 8-10 minutes of plane train time on average.

Let's do the math. Let's assume the plane arrives exactly at A0. We're going to use a "middle case" scenario here where a passenger is in the middle of the an average mainline plane (let's say a 738) with a statistically average concourse walk and transfer:

A0: Parking brake is set
A2: Boarding door is open (Airbus and Boeing flight ops planning says to assume up to 3 minutes, but we'll give benefit of the doubt to Delta for running a tight operation and getting the door open in 2 minutes)
A8: Passenger in the middle of the plane (seat 80 out of 160) gets off the plane - based on a deplaning rate of 15 people per minute per door (this is from industry literature prior to the removal of checked bags - in my experience, deplaning time has gone UP since passengers have more bags - but will use this estimate since not all flights are 100% full)
A11: Passenger starts waiting for the plane train (assumes the flight gets into a gate in the middle of one of the North/South splines - it is a 600 feet walk to the plane train, about 200 feet to get from the center concourse down to the plane train, and 50 feet on the jet bridge - and assumes the passenger is walking at 5 feet per second or 3.5 miles per hour - a pretty brisk walk, especially with bags)
A12: Passenger waits an average of one minute for the Plane Train (arrives about every 2 minutes at peak time - longer during off hours)
A16: Passenger gets off of the plane train two concourses down (e.g., A to C) as it takes approximately 2 minutes between stations when including wait times at stations
A19: Passenger arrives at the gate after an 800 foot walk (no jet bridge this time) - with 1 minute to spare before T-15 cutoff

If gate closure is done at T-15, then there is zero buffer for a flight arriving at A0 for the passenger sitting in the middle of a median mainline plane with the median transfer between concourses. Delta's A0 rate is about 60-65%, and there are some routes that are regularly under that. So I would expect that any route that is below 50% A0 arrival time (of which there are likely many) would have a misconnect rate at least close to 50% if booked at exactly MCT. Obviously passengers have some control (RUN vs. brisk walk) but it's not like it's some wildly off fanciful number.

There are multiple variables known at time of booking that would influence a likely misconnect rate. Those include: 1) the type of plane (739, A321, 753 all have longer deplaning times than the 160 seat 738 example I used), 2) the likely placement of the passenger in the cabin (a non-statused Basic Economy passenger is likely to be in the back quarter of the plane vs. a C+/First class passenger being at the front of the plane), and 3) the most common placement of the inbound and outbound flights - some flights are regularly scheduled to arrive at T, and some flights [e.g., regionals] are most likely to leave from D. It is not at all implausible to algorithmically know that certain types of tickets on certain routes have a predicted misconnect rate of greater than 50% if misconnect is defined as not making it to the gate by T-15.

This is consistent with what I have said: in practice, a 35 minute MCT relies on the assumption that the door doesn't really close at hubs until T-5 at the earliest. If there is a systematic effort to close doors earlier at T-15, certain tickets become essentially impossible to fly with any reliability (as defined by not missing a connection more than 50% of the time).

This should not be controversial, and the ridiculous numbers of "less than 5% misconnect rate for tickets booked at MCT" is just completely divorced from any reality.

edited to add: just to be abundantly clear, I am not saying ALL tickets booked at MCT have a 50% chance of misconnect; if I had to pull a number of our my rear end, I would bet that around 20% of tickets booked at exactly MCT misconnect at Atlanta in real world scenarios (including when doors ACTUALLY close); however, there are absolutely a portion of tickets that Delta sells that could reasonably be predicted by variables known at the time of booking to have a greater than a 50% misconnect rate IF misconnect is defined as arriving at the gate after T-15.