Apparently, Edge Canary allows keyboard navigation to disabled items when assistive technology like the NVDA screenreader is in use, but does not do so otherwise.

Consensus in bug 1702629 seemed to be that this may be worth emulating, but it is considerably more complex than some of the other options on how to deal with disabled items, so it is being split off.

I don't think we should differentiate behaviour based on the presence of a screen reader. If we're going to prevent keyboard navigation to disabled menu items, I think we should just do it across the board, accepting that we're intentionally inconsistent with native Win32 and accepting the potential user griping this might cause. There are two reasonable arguments in favour of this inconsistency:

1. From what I'm told, with the recent redesign, we're already deviating significantly from native Win32 visually.
2. It's impossible to get focus to any other disabled control, even native Win32 controls. In that sense, native Win32 menu items are the inconsistent exception here.

I think skipping disabled menuitem is no good ui.
Because, The muscle memories the position of the menuitem (how many times keyedown), so if it is skipped, it will change.

^ Then should we change toolbar key nav and everything else? Everything else I can think of already skips disabled buttons. Doesn't seem like it bothers people in practice or in principle. But the inconsistency is bothersome. You'd have to develop one muscle memory for menuitems, and an entirely different muscle memory for everything else.