Tracking updates missed from 20.20

In the process of investigating the remaining differences between 21.9 and 20.20, it was mentioned by @ncalace that there may have been some RungeKutta updates not yet included. @goblirsc found some plots showing that in 20.20 similar changes to what we observe in the difference appeared around May last year, at the same time as bundles containing changes from @gavrilen discussed in ATLITKSW-64 were included:

This ticket aims to collected any updates that may have been missed in 21.9 (some were already ported in different forms, but seemingly not all) in order to reproduce the final behaviour seen in 20.20. The changes so far affected downstream performance, but did not noticeably impact ID tracks including the number of strip holes in low-stats tests. Presumably the main changes responsible are still to be found.

adding also @asalzbur @npetters @tstreble @zschilla

UPDATE:

This ticket now collects remaining changes not in !28364 (merged)

These two together don't explicitly bring the strip holes into better agreement, but seem to result in a high cluster association efficiency, which is anyway a positive change. Currently checking the impact on Run 2 reconstruction to see that these changes don't have a negative impact there.

UPDATE 2:

From some checks, it seems that for ID tracks these change no output observables except for the covariance matrices themselves... for downstream objects using tracks however this does result in many differences as a consequence. Not clear how significant these are...

the fact that these changes remove some WARNINGs is perhaps a positive sign?

WARNING  Test log file run_q221/log.RAWtoESD has 7 less warnings than the reference log file /tmp/nstyles/clean_run_q221_7f912a3e-8dfc-41ea-b023-a2ce90e776bd/log.RAWtoESD
WARNING  The reduction of unnecessary WARNINGs is much appreciated. Is it expected?
WARNING  The following warning messages have been removed:
WARNING  ToolSvc.MuonTrackSummaryHelperTool                WARNING Number of hits in road < 0 in first projection: -4, setting = 0. (nhits in first projection = -4)

WARNING  ToolSvc.CombinedMuonTrackBuilder                  WARNING fit:: about to add the TSOS's describing calorimeter association to a combined muon

WARNING  ToolSvc.MuonTrackSummaryHelperTool                WARNING Number of hits in road < 0 in second projection: -4, setting = 0. (nhits in second projection = -4)

WARNING  ToolSvc.CombinedMuonTrackBuilder                     INFO  summary of WARNING occurences (maximum shown during job: 10):

WARNING  ToolSvc.CombinedMuonTrackBuilderFit                  INFO  summary of WARNING occurences (maximum shown during job: 10):

WARNING  ToolSvc.CombinedMuonTrackBuilderFit               WARNING fit:: about to add the TSOS's describing calorimeter association to a combined muon

UPDATE 3:

OK, according to @ncalace, the changes aside from those to TrkPatternParameters should already be in master anyway, and the TrkPatternParameters updates are purely technical and don't change output, so I think this is all OK to go in => removing WIP.