Mystery at Random Short Crossing

[OldEastRR]

After many hours of installing and troubleshooting my double-track double crossing, I ran into a mystery. Not the short -- I figured out where it was, in the insulfrog diamond, not the electrofrog diamond (yes, it was a dumb combination). This is DCC.
However, the odd thing is that the short only happens randomly. Some brand's engines ran through w/ no problem, some didn't. Some ran thru once then shorted out or vice versa. My LL DL109s shorted every time.
So I'm wondering how tread width, flange depth, wheell gauging, how the truck is engineered, slop in the axle openings, or other factors are involved in this. I'm assuming it's something like that since the quality-engineered KATO locos all passed without shorting, but other brands were hit and miss.
The short itself took forever to find because it was the most unlikely one -- somehow one of the metal guardrails at the center of the INSULFROG diamond is getting track power. And those things are supposedly dead rail. I have to disassemble my entire crossing junction to get at the problem so --- model railroading is so FUN!!!

[NtheBasement]

Never had that issue before and curious as to the results of your autopsy.  For the guard rail to get power either it is touching inside the plastic at the end near the point or it is touching the wire running underneath.  There are some open pockets in the plastic underneath, perhaps a piece of metal got in there?

[jjb62556]

I have had tread width problems on some of my switches where the gap of the rail at the frog is too close and tread bridges across both rails.I just painted with clear fingernail polish to cover a little more of the gap.

[OldEastRR]

What happened is some kind of conducting build-up (of what I have no idea) bridged the very narrow gap between one of the live rails and dead guard rail. so the guard rail shorted to the opposite charge rail when the loco's wheels crossed over the point where both rails were parallel. After I scraped all of it out of there the guard rail was dead again.
However my wiring design of having the two parallel tracks on a main have different polarities (one N rail carries the same charge as the other's S rail and same with the other rails) caused a much different problem. The Insulfrog diamond is wired with the assumption that both routes have identical polarities. So the crossing was mixing my two separate charges of DCC power in each rail. I had to basically rewire how the crossing rails and frogs are connected. Actually I so butchered up the diamond I need to buy another.
So it turns out I probably did the right thing in buying an Electrofrog diamond for the other crossing. Every rail and frog on that is isolated from all the other ones so no shorts can accidentally occur.