That behavior sounds very much like what I was seeing. My files were roughly an hour in length and I kept running into issues where it appeared as though my script was retrying the API after a failure, but that those failure responses ate into my credits. (Thankfully I was using free credits myself but it was a shame that it used them up.)
I don’t have the logs from the experiment but if I recall correctly, it took anywhere from half an hour to an hour to transcribe an hour of audio.
Hope that helps.