Body Ground - Frame or Engine Block?

I'm currently working through p0aa6 with info code 526. I'm getting a short between the body harness and the frame inside of the car, but after I got a techinfo subscription, it turned out that was jumping a few steps ahead. What I'm trying to figure out is what is considered body ground under the hood - the body frame, or the engine block?
I have a Fluke 1507 IR Tester set to 500V. When I test the inverter compressor connector to the frame, all three contacts return a resistance greater than the meter range. If I connect the ground lead to the engine block, all three contacts return a resistance value of about 2.5 Mohm, lower than the 3 Meg spec. I measured the low voltage resistance between different points on the frame and the engine block, and they are reporting > 22 Kohm. I know the short between the frame and the body harnesses goes away if I disconnect the frame wires from the inverter.

 

That's the thing, I got all of the HV leads out of the inverter (frame, MG1, MG2, A/C) and the inverter was still shorted to ground. I pulled the inverter and it smells burnt.

Also, I pulled the traction battery completely out of the car and the frame wires were showing a dead short to ground, which went away (at the frame of the car) when I pulled the frame wires from the inverter.

I actually suspected the battery first, since it's got 145k miles and I've had problems with the battery in the past.

I swapped the ATF and it didn't smell particularly out of the ordinary.

When I pulled the inverter, it smells burnt (I'm an EE and I get to do failure analysis on our products occasionally, I've gotten accustomed to the smell of roasted electronics), but I can't see any obvious scorch points. I haven't disassembled all the way to the IGBT's yet to be fair.

 

That would be a correct assumption. I actually pulled the HV battery, partly so it wouldn't be the cause of my headache, partly so it wouldn't be the cause of my untimely end.

 

The symptoms I was having were that I was initially getting an ABS/VSC/(!) light on the dash intermittently.
One day, warning changed to the red triangle of death.
I got my techstream lite out and pulled the code.
I was unlucky enough to get p0aa6/526 by itself, without the luxury of another info code to go with it.
The car was drivable after resetting the code (I read the Luscious Garage post after googling).
The triangle would set while the car was in neutral, causing me to suspect the battery.
Last week, the triangle came up while driving (not out of the ordinary at this point),
so I reset it using my BT OBD reader (Torque doesn't pull the HV ECU codes).
The triangle immediately came back, followed by a loss of power, the SOC dropping to one bar, and a burnt smell in my car. I managed to get off to the side of the road and have the car towed home.

As I mentioned above, I pulled the battery completely out of the car because it was my first suspect. Unfortunately, the short from the frame wires to the frame remained.
I've megged out every connector that goes into the inverter, and none of them have the ground fault. Once the MG1 and MG2 cables were off, I megged the U V and X contacts inside the inverter and they had dead shorts to ground. What I don't get is how that could be. I'm considering doing a failure analysis, but in the meantime I'm buying a used inverter.

 

OK, I put in a second inverter, and same problem. I decided to be a bit more diligent in probing for a fault. I disconnected all leads except for the frame wires, since they show a fault in the back of the car when connected to the inverter, but do not when d/c'd. The U V and X terminals inside the inverter showed the same.
I then disconnected the frame leads to ensure that wasn't causing the issue. Same dead short reading on U V X. At this point, the only points attaching the inverter to my car were the 3 bolts (two on the removable brackets and one on the non-transferred one in the back).
I decided to prop the inverter up a little so the back bracket wasn't touching the car. Sure enough, U V X no longer read as a short. The frame wires also no longer read as a short when reconnected. Where the third bracket sits, I saw some scratches in the black paint, and some kind of corrosion. I cleaned that off with some electronic cleaner, and the resistance to ground at that point increased, though was still below 10 Meg.
I'm going to lay up some Super 88 tape, since 3M specs it as 600V insulating, and -18C to 105C.

 

I was able to resolve the issue using my "faulty" test method. What I neglected to take note of was that the short only appeared when the return HV frame wire was connected to the inverter. The fact I was able to use some contact cleaner in the return lead made me start to suspect that the lead was shielded, and that there was a short between the shielding of the lead and its conductor. I think the contact cleaner was a red herring, it may have simply prevented the shield from temporarily making contact with body ground.

Since I had a second inverter handy, I took the HV frame harness connector off of my old inverter and inspected it. It turns out that the shield of the HV frame wire is attached to the inverter's housing, hence why the short in the back of the car to the frame wires went away when the inverter was out of the car. So to prove the issue was the HV frame wire return lead, I HiPot'd the connector shield to the negative terminal of the connector from the inverter at 500 V. I got >550 Mohm, so I felt confident that that connector was good. I then plugged the HV frame wire return lead to the connector and repeated the test, dead short @ 500 V.

I replaced the HV frame wire harness (lowering the suspension crossmember is fun!), and we're back in business. I ran the A/C on max cold for half an hour and drove 50 miles on the new harness and have yet to see a code return. My mileage is up to 47 MPG (partly due to winter finally mostly ending here in PA).

I might do a post-mortem on the old frame wire to see if I can figure out where things went bad, and do some speculation on what might have caused that, but I mostly wanted to post this to give some hope to folks who aren't (fortunate/unfortunate, take your pick) enough to get p0aa6/526 without any other code.