We use parallel tasks to speed up the npm builds in the pipeline. However we don't check the status of these tasks, so if one of hte tasks fails and so we have missing assets, this does not cause the pipeline to fail. It can happen in the production pipeline too because we build the image again there.

Example of the task failling for staging: https://app.circleci.com/pipelines/github/greenpeace/planet4-africa/784/workflows/5ae3b820-f7dd-4d34-99e3-1b66ff22bc06/jobs/3346. You can see in the backstop compare job that the pipeline just continued without the assets.

During the same deploy, there were actually multiple NROs that had the same issue. The issue already existed but it seems to have become a more frequent this deploy (probably because of general internet issues at the moment). We could investigate that issue, however we should assume that running "npm ci" can fail at any time, because it needs to fetch packages from a registry so it could run into network issues.

Definitely also a reason why we should build the image once, deploy it to staging and after checking deploy the same image to production. That would not solve this issue, but it would prevent us from deploying an image with the issue to production.