Negative reaction times audio task

OS (e.g. Win10): Win10 Pro
PsychoPy version (e.g. 1.84.x): 2022.2.5
Standard Standalone? (y/n) If not then what?: y

I got ocassionial negative response times when I piloted my experiment and I am having trouble understanding why or reproducing it. The task is to press space if the participant has heard the speaker before. There is a 1 second fixation cross, half of which has an ISI component.

The keyboard response component is in the next routine and starts at 0 with a duration of 1.5 (same as the audio). It is set to discard previous and record only the first keypress. Most of the reaction times are positive, but some are negative, and some are very negative (-1.48 seconds for example, which is too long to have been a response during the 1-second ISI.)

Could this be a pygame issue? I have problems with ptb on the lab computer.