Look for the fix, not everything
Judge the retest on the one thing you were working on. A single number wandering elsewhere is noise; the question is whether your priority moved in the right direction.
Give changes a couple of sessions to show up. One retest is a data point; a trend across a few is the real story.
Then decide
Improved? Move to the next priority. Looks fixed? Film one more clip the same way to confirm it. Some progress? Keep the same fix. No clear change yet? Adjust the drill or the cue before you abandon the fix.