Here's what it looks like to me. My *guess* here (I wouldn't call it an analysis without a LOT more info) is based on having taught basic and advanced formation for over 8 years in T-37 and T-6 aircraft and having seen almost that exact same situation happen hundreds of times with students.
It looks like the blue tail one (wingman) was inside the turn and overshot the red one, which appeared to be the lead aircraft setting up for a rejoin on the bomber. The blue one increased its bank as it overshot ahead of the lead, probably going blind as the lead acft went under his left wing. Very common problem with low wing aircraft in formation when they get too far forward and above the correct position. Due to the smaller turn radius being inside the turn, the overshoot situation got worse with the blue acft passing the red one on the inside of the turn while the wingman was blind. The pilot didn't do an unloaded low-overshoot (standard USAF procedure when overshooting inside a turn) or break out (standard procedure when going blind, aggressively pulling away from the last known position of the other aircraft), instead relaxed his pull probably as he was looking for the lead, and sagged into and in front of the red one. Looked like the prop of the red one clipped the tail of the blue one as it passed in front from inside the turn to outside the turn. You can see him immediately unload the G's and drop very low after the collision, possibly due to a dramatic change in pitch trim when the tail got chopped up. Damn lucky, you usually don't survive loss of the tail at low altitude since the plane generally pitches violently down when the tail comes off making bailout virtually impossible.
Normal turning rejoins like that usually have the wingman stacked slightly lower than lead, below lead's plane of motion. That way an overshoot is fixed by the wingman slightly relaxing back stick pressure and sliding below/behind the lead to the outside of the turn where he then flies a longer radius turn which solves the overshoot problem while keeping lead in sight. Pulling harder to the inside just makes it worse and leads quickly to going blind and having no options at all. If lead is diving or they're close to the ground though, its difficult and/or uncomfortable for the wingman to stay in that low position inside the turn. For that reason, low altitude turns are usually made away from the wingman whenever possible, not towards him, and the wingman is cleared to float to the left or right wing as necessary to remain aft of the lead.
My guess is the red tail pilot didn't see it coming and was fixated on his own rejoin. Keeping #2 in sight during multi-ship dissimilar formation maneuvering is difficult but very important, because your wingman might get jammed up inside a turn and have nowhere to go. In this case, the low altitude would make it very hard for the wingman to do a low overshoot and there may have been other reasons to not drop the left wing (instead of raising it) to keep lead in sight, or do some other vertical reposition. Pulling up and TOWARD the lead aircraft would help control the overshoot and minimize the time spent blind, while getting wingman up and out of the leader's plane of motion, and that's the last-ditch maneuver we would teach if the wingman gets jammed up fwd and high inside a turn. Pulling harder inside the turn just spits the wingman farther out in front, and at low speed the wingman may not be able to sustain a tighter turn than the lead so he'll eventually end up in a collision position again without any hope of regaining sight first because the other aircraft is below his belly outside of his turn.
If the flight lead expected his wingman to be outside the turn and not inside, he wouldn't have even been looking for him since you can't see someone below your belly outside the turn anyhow. That's something we can't know without a LOT more info on how the maneuver was briefed and executed.
Basically from the video, it looked like something I taught routinely and saw over and over (an over-shoot inside the turn followed by increasing the bank angle instead of doing an overshoot procedure or breaking out of formation), and I had to beat out of each and every student because the instinctive reaction is to do exactly what that guy did in the video, with predictable results.
I want to stress that the pilots may have had specific "be no" obligations that prevented a normal overshoot or breakout, or other factors that made a standard overshoot or breakout impossible. "Be no higher than the bomber". "Be no farther outside the turn than the lead or bomber". For example, he may have had a briefed contract to be inside the bomber's turn 100% of the time due to other aircraft outside the turn, the ground was close below him so a low overshoot may have been impossible, and again a contract obligation may have prevented him from climbing for a high breakout or high overshoot. So he may have simply had nowhere to go, which puts the burden back on the flight lead for inflight and preflight planning of a maneuver that hung his wingman out to dry. Also, while the flight lead shouldn't have to stare at his wingman to make sure he doesn't ram him, keeping #2 in sight is still a flight lead responsibility. Remember, this is my *guess* at what happened because I don't have all the info.
We had a somewhat similar midair at Sheppard AFB over a decade ago, where the wingman went blind and didn't break out right, the flight lead didn't see where he went, and they ended up in the same piece of sky. One plane landed all mashed up, the other crew had to bail out when their wing came off. Luckily they were at high altitude and had time to eject.
I've had a student put me in a similar situation as those mustangs, and I had tight altitude constraints too. I couldn't go up high due to other traffic and the flight lead was diving so I couldn't get below him. I ended up having to bank TOWARD the flight lead then push negative G's to bunt away while keeping him in sight. Uncomfortable to say the least, and that was my last pre-planned "way out" so if it hadn't of worked, we probably would have either collided or I'd have been a hazard to another plane or formation. I always tried to have 1 way out when I was flying, 2 ways out when a student was flying, and 3 ways out when I was teaching another IP. When training other IPs, I had to make a simulated student error (like pulling inside a turn during a rejoin), then if the instructor trainee made an improper correction, I had to have that second and third way out pre-planned just in case the new IP made the worst possible error trying to fix my initial simulated student error. Fun but challenging.
