Images and downloads no longer hang

Pictures, videos and file downloads stored on a node are served through the content route. On a portal running more than one replica, roughly half of those requests never answered — the page showed a broken image or an empty player, and a download simply stalled. Reloading sometimes "fixed" it, which made the problem look random rather than reproducible.

The request was being issued from the mesh's internal routing hub, so when the file's owning node happened to live on a different replica the answer had nowhere to come back to and the request sat until it timed out. It is now issued from a normal application hub, which every other part of the portal already used, so the answer always finds its way home.

Reconnecting…
The server was updated. Reloading the page to pick up the latest version.