power feeding through logic converter

[CraziFuzzy]

So I've started off with a relatively basic string of 12V WS2811 lights along the eves of my house, controller with a Pi4 I had laying around.  Pi is powered by a phone charger, lights from a decently sized 12V wall power supply.  Initially had some inconsistent performance due to trying 3.3v directly to the data pin, so breadboarded up a 74AHCT125.  Tapped the Vcc from the Pi's 5V pins on the GPIO header, which should be connected to the USB's power pins.  This actually worked great for three days, then on day three, the first pixel is glowing about 50%, and nothing else happening.  Pi is still functioning properly - FPP accessible from the page, and no voltage warnings.  Unplugged the USB to reset it to start my troubleshooting, and the Pi didn't power down.  I haven't had a chance tonight to start poking through it with a DMM (or possible scope), but somehow the pi is getting powered from the 12V through the logic converter circuit, and back through the 5V wire.  All I did so far was verify the breadboard is wired up right.

Anyone seen something like this that might shorten my troubleshooting?  Hate to have to to up to the eves and start half-splitting the strings to find a short or something.

[AAH]

If you are connected like it should be then there should be no logical way that the Pi could be powered by the logic converter. In theory you should have 0V and 5V coming from the Pi to the converter. You really should have a resistor (per output in use) between the buffer and the cabling out to your pixels. The resistor should be in the 22-47 ohm range though some controller manufacturers have used outside of this range. From the converter the 0V and data should go to some point where they combine with the 12V (   ) that is powering your pixels. That junction between the 0V and data and 0V and 12V could be adjacent to the Pi or adjacent to your first pixel. IF you have a resistor on the output/s of the buffer then even if the data input of the first pixel had failed and you end up with a short circuit there isn't a low enough resistance for the 1/2 amp'ish to power the Pi. If you don't have a resistor in the line then the failure mode would have to be something like 12V getting onto the data line, the internal voltage clamps on the output of the buffer clamping the "12V" coming down the line to the VCC rail of the buffer which is then connected to the 5V on the Pi.
Before you do anything I would make sure you have a resistor inline. To isolate and diagnose what is happening I would disconnect the connection between the buffer and the pixel and see if power is still being leaked somewhere.