So ultimately you can get stronger by getting more efficient via the nervous system or adding muscle cross-sectional area. Basically what a hugely arched bench does is limit the muscle that can contribute to the lift, which is in contrast to factor 2 above. The pecs produce the biggest force vector in a slight decline, or what you would get out of a minimal arch. Anything past that point begins to take the upper part of the pecs with its fibers that run directly across the chest out of the lift. This, in turn, removes a big contributor to the force production against gravity. From a holistic standpoint, you might get a short term bump from a big arch, but you have much greater potential in the long term by flattening out to a degree and optimizing the involvement of your prime movers. The pecs are the biggest muscle group in the lift and it's downright silly to not get the most out of them. Furthermore, you can always add more muscle to the prime movers and get stronger that way in addition to learning how to improve muscle fiber recruitment via technique. So it's the best of both worlds.
As for the lats, the short version is they just aren't in an anatomical position to contribute much. The common sciency argument is that when the elbows move past the midline the lats can pull them back to the midline (up in the bench) via contraction. While this is true, it would only really be relevant to someone with a huge arch. Most everyone doesn't have the necessary elbow travel for this to be a factor. And the reason for that is the issue of length/tension. Muscles have the greatest tension when they are the most lengthened, and the lats are almost fully contracted in the scenario above and thus have very little room to apply force. In addition, the lack of tension, stretch reflex and potential energy that are present in the prime movers (due to the fact that they are stretched out) further limits the lats. So you're looking at an almost fully contracted muscle that already had a very limited potential to begin with, so it's just not going to be very much of a factor. Furthermore, if it actually was doing much of anything in an arched bench, the elbows would be coming into the body initially as the bar moved off the chest, and I don't think I've ever seen that.
Lost in the lat argument is the very real importance of the traps. No one really talks about them, but the traps are the main mover of the scapulae. Everyone has heard the cue "pull your shoulder blades together," and this is done by the traps and rhomboids. Pulling the scapulae down is the lower trap primarily with an assist from the lats, but they just cross the scapulae and aren't the primary stabilizer. So essentially all the cues we hear for the eccentric are being executed by the traps first and foremost in guiding bar bath while the prime movers control the speed of descent.
I started to develop this philosophy when I noticed all of the old middleweight records seemed to be from guys with minimal arches, and none of the current huge benchers in the raw division have excessive arches that I can think of either. Additionally, I collected some pilot data for thesis research trying to determine lat activation, and it just wasn't there.
For a big bench, optimize the involvement of the pecs, anterior delts and triceps. Get them bigger with volume work, and stop worrying so much about the tiny contribution of the lats.
My favorite cue is to "reach the chest to the bar," which really gets my upper back to fire correctly. You'll notice Chest cave when the upper back is too weak.
--->post I made in Facebook strength crew. Will add to it, but it's a good synopsis of my feelings on bench.