I'm not doing this intentionally. Below are just my findings this year when searching for issues/PRs and reading other's code. So not sure how serious the problem is.

#4964 might be closed.
It is fixed in #6841
#5839 might be closed.
It is fixed in #6841
#7861 might be closed.
It is fixed in #8071
#7794 might be closed.
It is fixed in #9188
#9821 might be closed.
It is fixed in #9847
#6752 might be closed.
It is fixed in #8066
#5588 might be closed.
It is fixed in #8066
#3506 might be closed.
It is fixed in #8066

In the long run, I think it might be important to have someone check all the issues and PRs, especially some old issues/PRs. Compared with similar projects like numpy/scipy/pandas/matplotlib/tensorflow (many of them just 100+ PRs), seems that scikit-learn has to many PRs (nearly 600).