Now that the deprecation has happened, do we have any interest in deprecating linalg.interpolative so that we can switch to a more consistent and thorough randomized linear algebra routines?
This has been discussed briefly before in Add LSRN solver and the references therein. But I am not sure if there is any appetite for it. The reasoning is similar to what @rkern gave above. But in this particular case, it seems like the Fortran code forced our hand to have the suboptimal API in that shape.
Also in hindsight, randomized linalg moved on to become even more powerful and established since then. So I just wanted to entertain the idea since this might be the good time to do it if we want to do it, instead of piecemeal deprecations and creating an impression that every version we are removing some stuff randomly.