What it does

Three outcomes, not one: broken (404, 410, a 5xx a visitor would see), could not verify (refused or timed out), and works but sends your traffic off your domain. Only the first sets a non-zero exit code.
Every unique URL is checked once however many pages point at it, and every one of those pages is remembered, so the report says where to go and fix it.
HEAD first, GET when HEAD is refused - plenty of servers answer 405 to HEAD while serving GET perfectly.
A 404 on a domain's root is never a deleted page: crates.io answers 404 to anything that is not a browser, and the first version duly reported the Rust homepage as linking to a dead site. Both rules that fixed it are covered by tests.

How it is proved

Six pages of rust-lang.org, 220 unique links checked in 10.5 seconds at 24 concurrent requests. 28 tests, none of which touch the internet.

Rustsingle binary28 testsCSV output

People find this looking for

broken link checker no false positives, link checker ignores cloudflare 403, rust site crawler 404 report, linkedin 999 not broken, CI broken link check.

If that is your week and you would rather someone else did it, write to hello@dkautomation.dev or open an issue.