I decided to see whether I could reproduce the findings of varying +/- charge ratios as a function of p_T in Run 17 pp510 data in hopes of then searching for the source of the fluctuations. To that end, I examined a random sampling of ~30-40 MuDsts from 6 days of data: 075, 076, 105, 106, 140, 141. Here are my findings.

+/- ratio vs. p_T for all globals (MuDsts), all primaries (MuDsts), those primaries from the highest ranked primary vertex (PicoDsts I converted from MuDsts), and those primaries from any vertex matched to a BTOF hit:

day	all globals pT	all primaries pT	first primaries pT	btof-matched primaries pT
075
076
105
106
140
141

This is good in that in the primaries I can see considerable day-to-day fluctuations as well (globals seem to hide it), though I don't see quite the dramatic day 105 vs. 106 difference seen in the earlier study. However, days 075 and 076 go quite high! And curiously, requiring a BTOF match doesn't appear to suppress it much (which seems to contrast against the minor charge-splitting seen in TOF), nor does using primaries only from the highest ranked vertex.

Cutting on tracks with p_T>5 GeV/c, here are some plots of the +/- ratio for primaries vs. z of the primary vertex, track η, and track φ:

day	pT>5 GeV/c primaries Zvtx	pT>5 GeV/c primaries η	pT>5 GeV/c primaries φ
075
076
105
106
140
141

The evidence above suggests that the high charge ratios come mostly from a region of negative Z and phi just below 0. This is precisely where TPC sector 20 is!

In an effort to find out if this was more clear in a 2D plot, I tried to make a 2D plot of the ratio of primaries vs. φ and vertex z for all primaries with p_T>5 GeV/c. While it does still implicate the sector 20 region, the statistics are a bit weak on this. I'll re-show the p_T dependence of all primaries next to it so you can match the observables:

day	all primaries φ vs. Zvtx	all primaries pT
075
076
105
106
140
141

Conclusions at this point are not definitive, but TPC Sector 20 is implicated. There may be a good reason to suspect this sector because the "3D" GridLeak distortion correction was used for the physics production instead of the "Full" correction. TPC sector 20 had non-standard anode HV on channel 5, which would have suppressed the GridLeak distortion, and a standard distortion correction would be an over-correction. The "Full" correction should do a better job on that because it accounts for the anode voltages. Additionally, this over-correction would have a luminosity dependence. [UPDATE: I made an early look at using "Full" in the comment below]

-Gene