GitLab

As pointed out by Paul Gessinger, https://github.com/acts-project/acts/issues/1106#issuecomment-989624406:

One thing I noticed when looking at the MR over on your Gitlab is that you copied the pattern from our fitting algorithm, where I introduced an std::list to have stable pointers for the source links I create. This is not necessarily a good idea for a setup meant for any kind of production: std::list is relatively slow. I didn't add anything more clever to that algorithm there because it wasn't needed, but I would suggest you try to work around using this pattern, for example by creating the source links in one pass and only afterwards creating the measurements (when source link pointers are stable)

This needs some thinking.