Many conferences in my field (theoretical CS) are organised as follows. They are run by academics, as an endowement, scientific association, or some other not-for-profit organization. Each year, submission and reviewing are done on some platform like Easychair, producing a list of accepted papers. The authors of these papers are then requested to submit a final version of their work (or camera-ready) for inclusion in the proceedings of the conference. The proceedings are managed by a for-profit scientific publisher (e.g., Springer, Sheridan Communications, etc.), and the authors are asked to sign a copyright transfer agreement, so that the publisher can sell the final version on their website. The publisher may edit the final version, but in most cases it doesn't: the publisher's only role in the conference is apparently to compile these proceedings.
To make papers more broadly available, it seems to me that authors could prefer to simply host the final version of their work on an open repository like arXiv, as they often do anyway, instead of publishing their work in the conference proceedings. (Alternatively, they could simply submit the title and abstract to the publisher, along with a pointer to the arXiv version.) This would save authors the effort of dealing with the publisher's formatting requirement, and would also avoid any potential legal issues with the copyright transfer (which do not always allow authors to host their work elsewhere). Further, it seems to me that it would make no difference to all the rest of the conference organization.
Hence my question: Are some conferences OK with authors hosting their work somewhere else than in the publisher-run proceedings? More generally, do conferences care whether the authors of an accepted paper actually submit a camera-ready version to the publisher? If yes, why, and what happens when authors fail to obey? If no, why do authors bother?
(I imagine that authors may want to include the accepted paper to their résumé, but instead of listing their paper as being "published in the Proceedings of XYZ", they could point to the arXiv version with a note "Accepted/presented at the XYZ conference": presumably, this should make no difference?)
(Related: this question suggest that some conferences allow what I propose (for different reasons, apparently), but I have never heard about such conferences in my subfield. Why aren't there more of them?)