In late 2015, we were trying to establish Stv tracking in the forward direction, develop a (quick and dirty) fast simulator for the FTS Si tracker, and demonstrate the performace of a conceptual detector.  By December 2015 Stv was reasonably performant in the forward direction (for constant B field, of course)... and we were successfully reconstructing tracks in the forward direction at reasonable track densities.  The attached presentation served as my notebook during this time.

Slides 10 and 14 show the issue... several tracks were failing to be fit, yielding a curvature of zero.  Slides 11 and 15 illustrate these failures were occuring regardless of track multiplicity or radial segmentation.

The thinking was that tracks were passing through the same phi strip in each plane, and therefore ended up reconstructing with curvature==0.

To test this hypothesis we introduced a rastering scheme.  Each disk intentionally misaligned by ~1mm from the center, but at a different angle.  But the takeaway isn't that this misaligned the disks.  The takeaway was that we introduced a stereo angle between the phi strips between each plane. 

With that introduction, the degenerate fits disappeared.