Say you have an orthogonal matrix, you perform a few operations on it that are supposed to preserve it orthogonal property, but because of numerical errors you need to re-orthogonalize it. In this case, we can do faster than Gramm-Schmidt or QR as detailed in this SO answer: https://stackoverflow.com/a/23082112/1641621
This is also related to bug 1625
We also had this discussion about re-normalizing unit Quaternions (or unit-vectors in general) at some point (I can't find the reference to that).
The linked answer does assume that the input still is "somewhat" orthogonal, which must be clear from the documentation (and the question is whether we should verify this at runtime (in DEBUG mode). See also: Bug 601)
-- GitLab Migration Automatic Message --
This bug has been migrated to gitlab.com's GitLab instance and has been closed from further activity.
You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.com/libeigen/eigen/issues/1626.