is this fixing a known issue? can we have a test? How did you spot the pb?

it's claimed as default value of 0.1 in comment, and also in caller function, so the default value of 0.1 is the expected behavior.
I don't any existing bug because of that at this moment, since it's only used in gradient_boosting.py, and already taken care by caller's default value.
It should be fixed to avoid potential bug in future.

I think this is good and a typo fix. @pprett?
LGTM

So the alternative is to update the documentation to match the value, and hence not affect user code. If we update the value to match the documentation, we should at least note this clearly in What's New, as it's a breach of backwards compatibility.

I don't thing this will affect any user code, unless then "manually" called update_terminal_regions. Which I guess is technically public API.

This would atleast need a whatsnew entry... Otherwise LGTM...

Yes, please add a bug fix entry to whats_new.rst, then we can merge.

@glemaitre could you have a look at this? Apart for the missing what's new entry I think it might be good to merge with a +2.