The code used in
connectome2tck is not exposed anywhere else in the software. It’s an exceptionally simplistic approach, and also intrinsically depends on having access to the parcellation used in generating the connectome.
Generating a single exemplar trajectory from a set of streamlines (even for streamlines corresponding to a single tightly-coherent bundle, let alone bundles with more complex structure, or with data from multiple subjects) is a deceptively difficult problem. I had quite some difficulty with it some 10 years ago during my PhD. More recently there’s been some papers that have included details on how they overcame this challenge; I don’t remember exactly which papers they were, but if you do a search through recent manuscripts related to streamline-based statistics / registration you might be able to find one or more examples.
If you can find something with good evidence that it’s a robust algorithm and has enough detail to reproduce, it could be added to the list as a requested feature.