My suspicion is that this actually goes wrong somewhere in the platform-code, when we try to download a PDF document from a URL-string that does not (clearly) end with .pdf.

Until the PDF document has loaded, which in this case requires entering a password, downloading triggers the following code-paths through the viewer/platform code:

1. pdf.js/web/app.js

   Lines 998 to 1002 in c5dc082

   	} catch (reason) {
   	// When the PDF document isn't ready, or the PDF file is still
   	// downloading, simply download using the URL.
   	await this.downloadManager.downloadUrl(url, filename);
   	}

2. pdf.js/web/firefoxcom.js

   Lines 107 to 112 in c5dc082

   	downloadUrl(url, filename) {
   	FirefoxCom.request("download", {
   	originalUrl: url,
   	filename,
   	});
   	}

   where the url-parameter contains something like https://objects.githubusercontent.com/github-production-repository-file-5c1aeb/1663468/8963184?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAIWNJYAX4CSVEH53A%2F20220623%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20220623T123404Z&X-Amz-Expires=300&X-Amz-Signature=531edafe8105eb7743085ad1f806c9d1e8a39b2e5f1b9fb34284f4bb94f2f663&X-Amz-SignedHeaders=host&actor_id=2692120&key_id=0&repo_id=1663468&response-content-disposition=attachment%3Bfilename%3D05.-.Amplificador.diferencial.pdf&response-content-type=application%2Fpdf

3. https://searchfox.org/mozilla-central/rev/6d0ba065e3d41822337c708c8c0aca334ddd9218/toolkit/components/pdfjs/content/PdfStreamConverter.jsm#347-364

4. https://searchfox.org/mozilla-central/rev/6d0ba065e3d41822337c708c8c0aca334ddd9218/toolkit/components/pdfjs/content/PdfjsParent.jsm#100-117

5. https://searchfox.org/mozilla-central/rev/6d0ba065e3d41822337c708c8c0aca334ddd9218/toolkit/content/contentAreaUtils.js#54-72

6. My suspicion is thus that the following function isn't able to, in this particular case, determine that the URL-string actually points to a PDF document: https://searchfox.org/mozilla-central/rev/6d0ba065e3d41822337c708c8c0aca334ddd9218/toolkit/content/contentAreaUtils.js#194-263

Possibly a stupid suggestion, but: Maybe we could utilize the newly added aOriginalURL-parameter to help "inform" internalSave that this is actually a PDF document?

/cc @calixteman

Thank you for the in-depth look! I'm not at all familiar with web development or JavaScript, so I'm not sure I can help much.

I just tested in nightly, beta and release. I can reproduce the issue in either beta and release but not in nightly.
:ericonr, could you try on nightly to see if it is ok for you too ?

https://www.mozilla.org/en-US/firefox/channel/desktop/