I have written more extensively on these processes and the national airspace system management program of the FAA and DOT. Perhaps someone can find some of those discussions and cross-link.

United, and the other SFO airlines, have reduced schedules significantly during construction period. I tried to find a few reference links but they were all coming up with paywalls. The reduced schedules work fairly well when the weather permits but bad weather, as always, has a large impact on even these reduced SFO operations.

When, for whatever reason, an airline can not complete its full schedule it will manage its available resources (airplanes, crews, slots) to minimize the overall impact of the disruption. That means that some flights have minimal delays while others have large delays or are cancelled.

A simple first-in-first-out type approach would not produce a better result as the available resources would not be used as efficiently as possible.

The people who are making these decisions have to sort through large volumes of data, data that often changes after the decisions are make. The data comes from multiple system covering passengers, aircraft requirements, changing weather, crew schedules, legalities, and more. Automation systems are improving and providing those employees with better data to make better decisions. It's far from perfect, but the result is better than if the disruption was not actively managed. I think this is an area where A.I. has the protentional to make significant improvements.

The bottom line is that when an ATC restriction requires a delay or cancellation, whichever flight is ultimately delayed or cancelled will be attributed to the ATC restriction because, without the ATC restriction, the disruption would not have occurred.

This is the FAA's National Airspace System Status website. It is an interesting look into what is going on behind the scenes, particularly the Full Operational Plan page. https://nasstatus.faa.gov/