The stochastic nature of the conductive filaments in oxide-based resistive memory (RRAM) represents a sizeable impediment to commercialization. As such, program-verify methodologies are highly alluring. However, it was recently shown that program-verify methods are unworkable due to strong resistance state relaxation after SET/RESET programming. In this paper, we demonstrate that resistance state relaxation is not the main culprit. Instead, it is fluctuation-induced false-reading (triggering) that defeats the program-verify method, producing a large distribution tail immediately after programming. The fluctuation impact on the verify mechanism has serious implications on the overall write/erase speed of RRAM.