pages:deploy keeps failing for no reason

Summary

The pages:deploy job fails sometimes quite consistently for no reason. It says the artifacts are missing, but they were generated and uploaded successfully by the pages job, so if they are missing it is only because pages:deploy deleted them itself. The artifact is neither too large nor it contains many files. Retrying the pages job just causes pages:deploy to fail again, or hopefully some time it will succeed.

This is in gitlab.com.

Steps to reproduce

I wish I knew, it just happens "randomly" (a pipeline that succeeded just fails after a new push).

What is the current bug behavior?

pages:deploy fails for reasons only known to itself, and gives no feedback other than a short (and probably wrong) "tooltip" message in some pipeline views.

What is the expected correct behavior?

pages:deploy should succeed, or give a good reason for failing ("server error, try again later" could be a good enough reason), and provide some means of retrying without having to run the pages job again. In the latter case, pages:deploy should not remove the artifacts if it didn't succeed!

Relevant logs and/or screenshots

Output of checks

This bug happens on GitLab.com