chasing a misfire

Thought some people might be interested in my 06 4wd 165k misfire problem, and maybe in the end if it gets solved might help some people. This vehicle has had a random misfire apparently for years, and I got a few here and there, but it has progressively gotten worse. Now in cold weather (western NY), it's not random anymore.

Symptoms: on startup in cold weather I will get a flashing check engine light. Eventually once the engine warms up, it will stop flashing, and sometimes even the CEL will go off. (this would happen periodically when the engine was warm as well, but randomly) Random misfire, cylinders 4 and 6 and 1 are most common, but I think eventually all of them misfire. (but you know that could mean only 1 cylinder is misfiring, or all of them with honda) This vehicle I bought from a friend of the family a year ago or so, I got some history but feel like a lot is still unknown. I tried a few basic things first, checking spark plug tightness, inspecting spark plugs, looking down the cylinder with a borescope camera, nothing really jumped out at me except that I was dealing with a variety of part manufacturers. Some of the work below was done periodically, but as of today I am now focused on this misfire and solving it.

So here is what has been done so far:
1) all new spark plugs, ngk iridiums from rockauto. I pulled out 3 bosch in the back and 3 ngk in the front. They didn't look terrible.
2) all new denso coils(rockauto). The old ones I pulled out, some had nothing to indicate mfr on them (oem?), and there was 1 denso.
1 and 2 I did just to give myself a baseline, and rule those two things out. This car is also going to be my son's college car so I figured this was basically insurance money so to speak to keep him from breaking down less.
3) Timing belt kit with water pump (Asin, rockauto). 165,000 miles and still had the original belt in it!

Still getting misfires now, next thing I tried was a cylinder drop test by unplugging the coil packs, and the engine seemed to stumble on every coil unplug, so that pretty much ruled out a dead cylinder.

4) Adjusted the valves:
My first time adjusting valves, although through the years I've gapped many spark plugs so I'm familiar with feeler gauges. I wanted to adjust them right after the timing belt kit, but could not get to it yet. Took my time, double and triple checked with the proper size gauge, and the next size gauge up to make sure it wouldn't fit. Put it all back together, started her up, and get misfires! Took a few days to think this over, and decide to go back in and redo my own valve job. It was my first time, it was 15 degrees out (although warmer in garage but not by a whole lot lol) so thought maybe because not at room temperature too tight? (in hindsight I don't think so)
So I go back in, this time with 3 feeler gauges for each valve. And when I finished a bank, I rotated the engine several times and rechecked the valves. I checked and checked and checked, I wanted to be positive that my work is good, because as you know a lot of stuff has to be removed to do this job. (I also redid spark plug seals, coil seals, and valve cover gaskets) So after all this, I feel very comfortable doing valves and learned a lot of tricks to it. At least I have that now lol.

So now, sit in car, and start her up. It sounds like a brand new engine, just purrs along for the first 30 seconds or so. I'm thinking problem fixed. But soon, I can hear it start to misfiring, and it's very obvious listening at the exhaust pipe. Sure enough, CEL starts flashing and the same misfire codes will be generated. Once the engine warms up the misfires get better. One thing I did not know about hondas, they go from open loop to closed loop very quickly, like a minute? I was used to my trailblazer that doesn't go into closed loop until 5 minutes or so. But when I pulled freeze frame data I noticed the honda engine was already in closed loop mode. So this is telling me I might be looking at a sensor somewhere bad. Something the engine uses when it's in closed loop mode.

By the way, at some point during this whole process, I'm pulling scan data and looking at fuel trims, resetting the computer and codes, and pulling trims again. Bank one (closest to the firewall with cylinder 1) is pretty consistent about 1.08. So this means Bank 1 upstream AF sensor is reading a little too much air, maybe a mild vacuum leak. So less than 10% on a 17 year old engine, that looks perfect to me lol. But bank 2 clearly has a problem, it's consistently 0.87 or worse, I've seen it dip down to .74 at times. So Bank 2 AF Sensor is seeing too much fuel, and telling computer to reduce fuel for bank 2. So maybe a leaky injector. Because only bank 2 is reporting a problem I think we can rule out anything that affects both banks like fuel pressure, intake issues etc.

5) cylinder leak down test:
Honestly I should have done this first thing when I thought about buying this vehicle, or at least a compression test. But I did drive it, it seemed to drive ok, and figured whatever is wrong with this honda I can fix lol. So I took off radiator cap, oil fill cap, oil dipstick, pcv hose end and loosened the throttle body to give it a gap for air to escape at the gasket, and removed all the spark plugs. Instructions say to do the leak down with engine warm, but I did mine with the engine cold, because that's when the misfires are happening. Tested each cylinder at top dead center (I used a borescope camera to help read what cylinder it's on at the front top timing cover viewing hole). Everything tested great, I was actually impressed. All cylinders under 10% leaks, and all the same measurements, any air I could detect was coming out of the crankcase, probably due to worn rings I think. But even then under 10% (just under), I think this is a pass. No bubbles in radiator, no air coming out of throttle body, no air at exhaust pipes. I did on purpose rotate then engine some, first to open up exhaust and it was easy to feel the air coming out of the exhaust with my ear. Then I rotated the engine to open up the intake valves and it was easy to tell that as well. So nice results from leak down test.

However, during this process, I did see this with the camera down cylinder 4:


It's wet, but with what? With negative fuel trim on bank 2 my first thought was leaky injector. But it could also be oil or even coolant. Hopefully leak down test ruled out coolant. How do you tell what this is? Normal qtip not even close to being long enough, plus it's off to the side of the cylinder. I tried sticking a plastic cable strap down there, and thought I got tiny little bits and pieces on it, but it would not ignite when a lighter was put to the cable strap. But this is after sitting all night so not sure if it's fuel or oil. With a negative fuel trim my first thought was to run out and buy a fuel injector. But before doing this I wanted to get more scan data and verify my AF sensor was working ok, because that whole closed loop thing is bugging me.
Here is what I am seeing:


Look at bank 2 short term trim, vs bank 2 downstream sensor, and how it compares to bank 1. See how bank 2 AF sensor is seeing extra fuel, and telling the computer to use less fuel, but the downstream sensor is running very lean (downstream sensor is seeing extra oxygen) ?
I think the bank 2 AF sensor is lying. So I started doing some accelerator snap tests, and no matter what I did, I could not get bank 2 sensor to go lean at all above 1. On a few occasions, I did get it to run richer (drop down to 0.75), and I noticed the downstream sensor go up in voltage indicating a more rich environment.

So the first thing I'm going to change I think is the bank 2 AF sensor. I may get some more data at startup today, maybe pull a vacuum line or inject some propane to see what happens (it will affect both). But I'm pretty sure that sensor is bad. Interesting is that I've never gotten any type engine codes other than misfires. Ok that's it for now, I will update no matter how this ends. I may put a new injector on cylinder 4 anyway to rule that out. By the way, here is some freeze frame data: