The logic is that it takes the collaboration to operate the detector and all the myriad of tasks needed to process the data to be ready for the final stages of analysis (where smaller groups do the work that is mainly described in the paper itself). 2/n

One could imagine many different models. Eg. for designing and building the detector you could have very highly cited detector papers with a smaller group that is responsible for that. 3/n

you could imagine some opt-in approach, but that is clearly going to be gamed. A manual process of deciding who is an author on a particular paper would be complicated given 3000 authors.

The unfortunate result is that there is very little information conveyed in the authorship. That's not helpful for hiring and it's not good for motivation.

In astrophysics my understanding is that often (eg. Fermi-LAT) that core contributors to the instrument have ability to opt-in on any paper written by the collaboration, but the members of the collaboration are allowed to write limited authorship papers.

Recently, ATLAS and CMS have experimented with some limited limited-authorship papers for specific types of studies (eg. demonstrations of machine learning on synthetic data or similar special topics). This is partially in response to dissatisfaction in monolithic approach

It's clear that scientists want outlets to express themselves in a way where attribution is more obvious.

It is true that for faculty hiring and tenure & promotion committees the current approach to authorship can cause problems (at least in the US). It is true that the role of letters becomes more important as well as holding leadership positions (eg. convening a working group)

Those give a fuller picture of a persons role, but it has it's own problems. It's not the best for increasing diversity and management / leadership roles don't necessarily align with what makes one a good & impactful physicist.