I am still seeing bad tracks coming out of the tracker recon in MAUS v2.8.2. Using Durga's log file, I get e.g.

...
Cov nan in run 9153 spill 24 event 12
     [nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan]
...
TP nan in run 9153 spill 809 event 26 : {'pz': inf, 'py': inf, 'mass': 105.6583668, 'y': -1.8499726139571353, 'x': 2.5217605057389583, 'px': -inf, 'z': 15068.860691319645, 'pid': -13}
...

where Cov nan means a nan in the Kalman covariance matrix and TP nan means usually an inf in the track point vector. I attach a full log file with all of the buggy entries, from the first 1000 spills of run 9153 reco pulled from here