When I was working in #824, I found what I think is an issue. But there is this test which was failing if I tried to "fix" it. So I thought it was not an issue so I didn't apply a fix, but I would like to discuss it.

I will create to PRs, I will ask you to decide which one we should apply and we can close the other one:

• Fix #844 alternative 1: remove incorrect bug notice #845 if you want to conserve the current behaviour
• Fix #844 alternative 2: do not allow access to unexported struct fields #846 if you agree with me