Why is the audio waveform vizualization in video sequencer not synced with audible audio while scrubbing?

I’m trying to measure latency between a certain audio event and visual event in a VR application. To do so, I record a 240fps video with a GoPro aimed at one of the lenses/displays of a Quest 3. I want to use the Blender (4.5) video sequence editor to count the amount of frames between a loud beep and a certain visual event shown on screen.

But I noticed there is quite a desync between the audible audio beep while scrubbing and the waveform visualization on the timeline. As indicated by the screenshot, the beep is audible on frame 114 while the visualization shows it’s playing around frame 120. So which is the truth? When did the loud beep actually play?

The offset of the audible beep does seem to sync well with the visualization.

I enabled “Sync to Audio” and “audio scrubbing”.