Checking TPC calibration with TOF in Run 17 pp510

I asked the TOF group about making plots that could help assess the state of TPC calibrations in Run 17 pp510. One of the helpful metrics used in the past is the fit the mean value of mass2 for each of the species π,K,p for + and - charges separately and see if there is a charge-splitting.

Zaochen kindly agreed to do this and the below are his plots (all credit to him!) for three sets of data:
(Note: you can click on the three plots to take you to Zaochen's original PDFs)

days m2 fits

The results look pretty reasonable, with perhaps a charge splitting at the 1-2% level for days 140+141. To understand whether we should expect anything visible, I went back to the studies of h-/h+ issues from a decade ago. There, we saw plots like these, which showed a charge splitting of several %:

The cause of the h-/h+ charge-splitting back then were TPC alignment issues that altered q*pT by ~0.002*pT2, or Δ(q*pT) ~ 0.002*pT. So distortion issues at the 0.002 level are very clearly visible. What should we expect now?

Back in 2017, we had an estimate of Δ(q*pT) ~ 0.005*pT for the Run 17 pp510 data in the context of the problem being a discrepancy between using the "3D" and "Full" GridLeak distortion correction models. With that, and using a power law with the form pT-2.83, it was shown that +/- ratio changes of the scale seen in the primary tracks in 2017 pp510 data were possible. That level of pT-shifting implies we should see a ~2.5x larger charge-splitting effect now than in the h-/h+ era if the choice of models is the reason. But Zaochen's results do not show something larger than in the h-/h+ era. That explanation was also dead-in-the-water when examined back in 2017 because it would require global tracks to show a larger effect than primaries, which is not the case for the Run 17 pp510 data.

Just to come up with another estimate and sticking with recent findings, I looked at Jae's recent plots. If I fit the spectra over pT = [6,10] GeV/c for his period 2, I find a rather reasonable fit to the data using the form e-0.452*pT. This may not be the appropriate parameterization, but it lets me get a handle on how much pT must be altered to see the variations Jae sees. I find that a change of Δ(q*pT) ~ 0.008*pT is necessary to alter the charge ratio from his period 1 result to his period 3 result. That implies an effect ~4x the size we had in the h-/h+ era. An even larger discrepancy that we should see clearly, but in fact don't see now in Zaochen's results!

After all this, the conclusion seems to be that the charge splitting in m2 from BTOF-matched tracks show a calibration in the TPC that isn't compatible with the large +/- ratio variations that Jae sees. My initial thought was that the problematic tracks may not be BTOF-matched (e.g. pile-up, or so distorted they don't project to their BTOF hit), but unfortunately I have seen that there's still evidence for the problem with the BTOF-matched primaries.