Well in various CS fields people preprint on arXiv these days, so it is easy for reviewers to unblind themselves. Sometimes accidentally, because preprint timing and subsequent Twitter advertising coincide so well with conference deadlines...
That would be a good idea, and then have the names revealed in a future version like the way the paper itself can be updated. I don't think it is possible though. If it is, it certainly isn't used.
The main point of arXiv is not to substitute for actual publication. Some people use it that way, which could still occur. But most intend to submit their work for review to a venue with a supposed double blind process. The use of arXiv is for flag planting and ease of open access.
Often the arXiv upload will be right around the time of a conference deadline in fact. For these people I see no problem with an OpenReview-esque setup where the arXiv page remains anonymous until the authors opt to reveal themselves.
If people don't care for true double blind review then they should stop pretending they do. Conferences like ICLR and NeurIPS are far from blind. It would be one thing if it were just a matter of trusting reviewers to not Google the paper title, but it goes beyond that. There are currently no restrictions on the use of social media to advertise works under review nor on the timing of arXiv uploads, which can trigger alerts to those in the field right around review time.
