Reference style links are checked or reported multiple times

[theofidry]

Taken from this job: https://github.com/box-project/box/actions/runs/13615816913/job/38058706641

And an excerpt in case the job output is removed later:

✔ [200] https://secure.php.net/manual/en/phar.setsignaturealgorithm.php
✔ [200] https://secure.php.net/manual/en/book.zlib.php
✔ [200] https://symfony.com/slack-invite
✔ [200] https://symfony.com/slack-invite
## Summary

| Status        | Count |
|---------------|-------|
| 🔍 Total      | 644   |
| ✅ Successful | 642   |
| ⏳ Timeouts   | 0     |
| 🔀 Redirected | 0     |
| 👻 Excluded   | 0     |
| ❓ Unknown    | 0     |
| 🚫 Errors     | 2     |

## Errors per input

### Errors in doc/docker.md

* [404] [https://docs.docker.com/samples/library/php/](https://docs.docker.com/samples/library/php/) | Failed: Network error: Not Found

### Errors in ./doc/docker.md

* [404] [https://docs.docker.com/samples/library/php/](https://docs.docker.com/samples/library/php/) | Failed: Network error: Not Found
*

You can see the 404 appears twice. The underlying code was something like this:

blablabla [`docker-php-ext-install`][docker-php-ext-install] blablabla [`docker-php-ext-install`][docker-php-ext-install]

[docker-php-ext-install]: https://deadlink.com

In other words I am using a reference-style link in my document and it looks like it is either checked or reported for every usage, rather than once.

[mre]

I think the problem is more that doc/docker.md and ./doc/docker.md are listed as separte documents.

When I search for docker.md inside the project, I can find a few links:

> rg 'docker.md'
./doc/symfony.md
104:« [Docker support](docker.md#docker-support) • [Reproducible build](reproducible-builds.md#reproducible-builds) »

./doc/code-isolation.md
75:« [Optimize your PHAR](optimizations.md#optimize-your-phar) • [Docker support](docker.md#docker-support) »

./README.md
27:- 🐳 [Docker support (`box docker`)](doc/docker.md#docker-support)
98:1. [Docker support](doc/docker.md#docker-support)

./doc/index.md
24:- 🐳 [Docker support (`box docker`)](docker.md#docker-support)

./mkdocs.yaml
50:            -   'Docker support': docker.md

And when I look at your lychee args, I can see a glob filter for 'doc/**/*.md':

--verbose --no-progress '*.md' 'doc/**/*.md' --cache --max-cache-age 1d .

lychee discovers files through command-line arguments (inputs) like these glob patterns, but also through links to other documents. That might explain why doc/docker.md and ./doc/docker.md get discovered.

We don't do any deduplication when listing the per-file summary and it's also pretty challenging due to many edge-cases, but perhaps we should give it a shot.