Possibility to get more reliable vocal RTs in online experiments

I have a more general question about implementing the microphone component in online studies. I would like to run an online study in Pavlovia that presents a sound file to the participant. The sound file contains a question that the participants are supposed to answer with ‘yes’ or ‘no’. What I’m interested in is the time between the auditory stimulus offset and the participant’s voice onset. To test timing accuracy I set the microphone to start recording one second before the stimulus sound offset and checked how much of the last second of sound it captured. I have found some variability in when the audio recording actually starts, compared to when the csv data file reports that it started. This is of course a big problem when measuring vocal reaction times (as has also been discussed in this forum).

I read here that the microphone.start component implements this via a promise, and that the fulfillment of the promise takes about 50-200ms. I was wondering if there is any way to log when exactly the promise has been fulfilled? Or is this a problem similar to the one discussed in this post?

I would be grateful for any input or other ideas for getting a more reliable estimate of when the recording actually starts. Thanks!