I have talked with Nathan and Rafo about this, and we'd like to get the SVT experts' opinion on it.  It seems that nobody is using the seed tracks for any of their analyses.  As it is, we have separate copies of each final state particle, vertex and moller for the GBL tracks versus for the seed tracks. I had thought that by making the seed tracks transient (ie, not stored in the output files), we would greatly reduce the file size by about 30% or so, but it turns out that it only lowers the output file sizes by 2%.  I am surprised this isn't a larger fraction and this is hardly a large enough change to justify removing it.  Are there other things associated with the seed tracks that could be made transient as well, such as residuals or HelicalTrackHits?