When provided with 3.3V and a fast-rise 5V input, this PRT-14765 Logic Level Converter oscillates on each positive cycle of the input at about 50 MHz. For example, with a 1 PPS input, the positive cycles of the 1 PPS get filled with a 50 MHz square wave on the 3.3V output side of the converter.
My guess is that the inherent potential instability of the TXB0104 bidirectional translator chip is responsible, although the addition of noise from the TPS61200 boost power chip may not be helping. I didn’t try switching to a 5V supply to see if that fixes the problem.
I recommend using uni-directional converters when feasible. I also recommend that Sparkfun produce similar products with a single supply which provide uni-directional level translation.