In a traditional TUI you can type at full speed and the system will process the keystrokes as it is able, with no keystrokes being lost. If a keystroke calls up a new screen, then the very next keystroke will apply to that screen. Competent users could be several screens ahead of the computer, because the system doesn't make them wait for the UI to appear before accepting the keystrokes that will apply to that UI.
It's not something windows does or can disallow. You just focus the window in the code and that's it.
Reminds me of https://en.wikipedia.org/wiki/Therac-25
What works best is heavily task dependent, and also needs to account for cognitive load and training requirements. Stations with few primary tasks taking limited input (e.g., product lookup forms and customer purchase forms) lend themselves heavily to a keyboard driven UI, but many others either do not. In that area, it ends up just being a strong personal preference guiding what seems best.
Keyboards are just faster once you learn them.