Thanks for checking!
tldr: That's
https://bugzilla.mozilla.org/show_bug.cgi?id=1388738 but it's
not a regression.
The difference we see here is the outcome of Bug 1388738, where we started
registering the Context Menu entry point (ie users that start interacting
with devtools using the "Inspect element" context menu entry). The category
existed before, but wasn't used.
We can see that the Context Menu category grew from 0 to 28% while the Key
Shortcut category fell from 65% to 38%.
This probe only registers the _first_ entry point used for a given session.
Before Bug 1388738, if a user did "Inspect element" and then switched to
another panel or closed devtools via a shortcut, a Key Shortcut
entry-point would have been registered. After Bug 1388738, we register a
Context Menu entry-point for this flow. So we just used to miscount 28% of
users as coming from a Keyboard Shortcut.