- with experienced users, e.g. people who are used to vim
- with a sharply limited number of modes, e.g. the three modes that vim has
It's just really easy to get carried away when you have the enormous number of combinations that touchscreens allow.
with experienced users, e.g. people who are used to vim
... though, is that it basically represents a surrender on the idea that UX is important at all. Given enough time and commitment on the part of the user, any interface can be learned well enough to be adequately useful. You could build an interface out of loaded guns and rotating knives, and a sufficiently committed user could eventually learn how to operate it without killing themselves. The challenge is that most users' time and commitment are not unlimited.
A stronger argument in favor of modal interfaces is that they enable the creation of complex interfaces that wouldn't be possible without modes. This is part of what appeals to so many vim users about vim -- once you've climbed the learning cliff it presents, you can do some things much more efficiently than you could in, say, a WYSIWYG editor. I personally don't agree with that tradeoff, but I recognize that a number of people see it as worthwhile.
Consumer-oriented UI is not the only UI! Users' time and commitment are not unlimited, but they are substantial when e.g. you are in the military and already need to be trained to understand the problem domain.
EDIT: A great example is the UI of the controls of a car. It's important to put commonly-used controls in ergonomic areas, follow the principle of least surprise, etc.; but it can still be built with the assumption that the user is a licensed driver and knows where the turn signal usually is, that up is left and down is right, what all the assorted symbolic labels mean, etc.