A file you just uploaded serves straight away

Uploading files kicks off background work — each new file is indexed so it becomes searchable and gets a description. While that was running, asking for one of the files you had just uploaded could fail: the page waited ten seconds and then reported the file as unavailable, over and over, for as long as the indexing took. An editor who uploaded a video and previewed it immediately saw a broken player for minutes, with no way to tell whether the upload had failed or was simply not ready — and every attempt cost another ten seconds. Then, untouched, it started working.

The two had nothing to do with each other except that they were queued behind one another: reading a file's location and writing the indexing results were being handled by the same single worker, in order, so a read had to wait for every write ahead of it. Reads now have a worker of their own. A file serves as soon as it is uploaded, whatever background work is still in progress — and the same applies to ordinary page loads during a bulk import or a package installation, which could previously report content as unavailable for the same reason.