Permissions consent prompts are broken. they are far too coarse, and out of context. The customer is forced to consent without knowing the scope or what will break.

A better experience would be an async consent screen, with placeholder content and calls to action until permission is granted.

For example, a network ping scanner could show placeholder host names & status until the user clicks to allow access.

The operating system could easily stub the sensitive APIs with placeholder content and a flag indicating the content is false. So even legacy apps could work.