pic '708 best' has one cell hot top left? One cell would make that string of say the three, weaker.
It looks like 708 (1) mid-left, (2) mid-right, (3) upper-left, (4) upper-mid all have cells that are running hot. From what you are saying these 4 hot cells could be reducing the output. That makes sense.
But I don't see any similar hot cells on panel 809 which had the worst current output of all my panels. That is puzzling.
I am still suspicious of the MC4 connectors, also. I found that the connectors were installed at the panel and inverter factories (not by the installers). That implies that the proper crimping tools were used and the connections should be good. If they were attached by the installers using pliers rather than a crimper the joint would be highly suspect. But that is not the case. I will inspect the connectors if the installer lifts the panels. I should be able to determine whether a crimper or pliers was used.
If my calculations are correct, a 1 ohm resistance in the connector would drop the current from 8.5A to 6.5A. The power dissipated in the connector would be about 40W. That would make the connector quite warm, but not blisteringly hot. When the installers lift the panels I will measure the temperature of the connectors with the thermal imager. With a near-zero resistance I expect their temps to be about the same as the shaded roof temperature. If the temperature is significantly above that the connections may be an issue.
The microinverters have been eliminated as the culprit because their efficiencies are all above 95% and unchanged from Day 1. Currently waiting for the panel manufacturer (LG) to get back to the installer with further instructions.