This is the Philly paper. You can explore through its cited bys and references if you want to see the continuing state of the research, but it's pretty rock-solid. There's very little doubt in the minds of any policy experts I know of or have read that signaled intersections, in urban contexts, should be used far less. That all-way stops are almost universally a safer design.
Your response on my points about delay is very much just one small problem thinking. I admit, LA's traffic situation is utterly fucked (thanks to putting the car at the center of all their urban planning for decades, which results in cities that are somehow undriveable AND impossible to navigate outside of cars at the same time). As a person who is immersed in this (and currently published in the TRB, if you can take my word on it because I won't be doxxing myself), let me assure you: traffic engineers are lazy, unimaginative fuckers. They follow their design manuals like bibles. ROR is easy to execute so they execute it rather than spending the extra 30-40 minutes to include more comprehensive phasing in their proposals. The manuals tell them that's all they have to and most others are too scared to challenge their "expertise".
Any traffic system that is going to gridlock because of removal of ROR was misdesigned. Period. Also was probably going to do it anyway, especially as traffic naturally grows over time (outside of the effective policy projects to reduce traffic, e.g., complete streets/multi-modal transportation plans).
If it is low enough volume that it makes sense to have ROR, it shouldn't have the signal at all.
If it is high enough volume that it risks serious problems if ROR is removed, the ROR almost certainly unsafe to begin with and a dedicated turn signal should be incorporated. Even if it just a signal indicating when it is acceptable to make an unprotected right on red.
ROR is currently the default and "opt-out" in relevant US intersections. It should, at best, be an opt in (e.g., with an arrow indicating you can turn right while yielding during certain phases).
I am not saying all traffic lights should go, but we have far, far, far too many of them. ESPECIALLY in the US, where they basically always have extremely simplistic phasing that, outside of peak rush hour times, simply increases average trip times.
To put it another another way: Braess's paradox hints at a larger truth: the systems that intuitively seem helpful to prevent congestion are often what CAUSED the congestion. There's no strong research on AB testing for congestion vs traffic signal removal that I am aware of, unfortunately, because the study is just laden with confounders eliminating any real AB comparison (e.g., making streets safer for multimodal traffic, e.g., by removing signals and replacing with all-way stops, leads to fewer people driving and that may be the "real" reason congestion goes down).
Don't miss the forest for the trees. Removing right on red is a safety win anywhere you do it. The congestion effects, if and when they even exist, can be addressed through separate system adjustments.
RE: crime... nothing is a better crime deterrent than humans present. My prescription is still to make the streets and neighborhoods more walkable. Adjust policies and designs to get more people comfortable being out there. Not even going to get into challenging the idea that crime is truly on the rise -- we both know that it isn't really.