I had the bitmasks for checking beamspot convergence set wrong since the start of run 3. Fortunately I don't think it should have caused any problems in the data we've taken so far, see ATR-29478 for details.
Thanks to @rowang for spotting this.