It's always DNS
I finally decided to add an admin dashboard to effortlessly dive into feed loading issues and pinpoint dead ones. The most striking discovery was a flood of generic connection errors questioning whether the machines were even reachable. When I tried loading a few with curl, none of them responded—the address resolution was failing miserably. So, I switched my server's DNS to well-known public ones and cleared out the failure logs, which brought most of those troublesome feeds back to life.
To make debugging similar issues easier in the future—for myself and anyone else who decides to host my platform—I added explicit DNS lookup in the code to provide more meaningful error messages than the generic ones from Axios.
It's wild how adjusting something as fundamental as DNS settings can resurrect so many feeds. Sometimes, the simplest tweaks unravel the most stubborn issues. Makes you think about the hidden bottlenecks that could be lurking in other systems, just waiting for a bit of attention to get things flowing smoothly again.