Unifying an invariant lifetime with a covariant one is almost always bad

What it does

Triggers on code of the form &'a mut Something<'a>. (The mut is important here.) I see people write this when they're not sure what to put in each lifetime position, but it's (as far as I can tell!) basically always wrong, and causes weird lifetime errors in other places by unifying the two lifetimes. This is particularly annoying when it appears as a function argument, as it causes weird errors in the caller.

Unless I've missed something this code should almost always be &'b mut Something<'a> instead.

You'd probably also want this to point out cases of &'a mut &'a Something.

Advantage

Avoids unifying the invariant lifetime of the mut reference with the (often covariant) lifetime of the (often not mut) reference inside the parameterized type.

More importantly, avoids distant confusing errors from rustc attempting to solve lifetime constraints in callers.

Drawbacks

It's possible that this pattern has a use and I've never seen it -- if it does, this would false positive on such code.

Example

&'a mut Something<'a>

Could be written as:

&'b mut Something<'a>

No files, tests, or entry points are named. Start by reviewing how Clippy implements lifetime-related lints, then assess the examples involving &'a mut Something<'a> and &'a mut &'a Something. Done means establishing the supported patterns and false-positive boundaries before adding the lint and its validation coverage.