A user reports a recurring issue with OneBox rendering links incorrectly on their Discourse forum, where links initially display properly but then show a broken link icon. The problem is observed with links from sites like neo.life and nature.com, with some links failing due to the site blocking bots (e.g., Cloudflare captcha), while others fail due to technical issues in Discourse’s dependencies, such as broken favicon handling. The community investigates, with one user identifying that the issue with neo.life is due to the site blocking image crawlers. A suggested fix is to add the domain to the disabled image download domains list, though this requires rebuilding the post’s HTML. Another issue is traced to a bug in a dependency, which is being addressed via a merged pull request. The team acknowledges the problem and requests more examples to further diagnose and resolve the issue. BCHK, darkpixlz, sam, david, Stephen, nbianca}
When I post this link in my forums after it first renders appropriately.
actual link below here: https://neo.life/2022/07/the-buck-institute-where-the-promise-of-aging-research-isnt-longevity/
but then a few seconds later I get get a broken link symbol, so, it looks like this in my forum:
No - it also happens here:
This link is here: https://neo.life/2022/07/the-buck-institute-where-the-promise-of-aging-research-isnt-longevity/
but then if I put it on its own line:
https://neo.life/2022/07/the-buck-institute-where-the-promise-of-aging-research-isnt-longevity/
Something is certainly off here, we will have a look. Maybe the host is rejecting the user agent for our backend crawler.
When accessing the image (https://neo.life/wp-content/uploads/2022/07/buck-institute-longevity-aging-genetics@2x-1000x714.jpg) programatically (e.g. via curl), we get served a Cloudflare error page with a captcha. So it looks like this site is deliberately blocking robots from downloading the images.
If you trust the site and are happy for images to be permanently hotlinked, then you could add neo.life to the disabled image download domains to stop Discourse attempting to download them.
Interesting… I tried to do what you suggested (adding neo.life to the disabled image download domains), then opened the post and re-saved it , but the problem is still evident.
Do I need the http:// in front of the domain for it to work properly?
I’m seeing this onebox issue on more and more links lately - another example:
Same thing happens - it renders, then it doesn’t.
To fix an existing post after changing the site setting, you’ll have to click the 
, then “Rebuild HTML”. I just did that on your post above and it’s looking ok here. Can you try that on your site?
This one looks like it has a slightly different cause. I’ve tracked it down a problem in one of Discourse’s dependencies. We’re tracking this, and will try to get a fix or workaround in place. Thanks for the report!
Another example - yes, the rebuild HTML worked in that last case:
But here is another example where I’m getting the broken link image for a respectable sourced link:
Even if the site is considered respectable it doesn’t mean that they’re going to allow images to be crawled and downloaded by bots.
Keep the examples coming @BCHK !
We recently fixed some issues oneboxing guardian, eg:
The one from MIT is happening cause they send through a favicon to a missing image. We plan to fix this by simply never rendering these broken chains in oneboxes.
Just noticed a PR was opened that seems to be related to this
You are correct. That PR is merged and you can see the dead favicon above is no longer showing up as a broken chain icon.
