In the deploy_k8s_branches job, the metadata for the data dump can be incorrectly formatted (when the stat command doesn't produce valid YAML, for example). This will cause the yq command to fail, but the ETAg gets overwritten anyway. As the ETag is overwritten despite import failure, any subsequent runs will skip the data import as there is no longer a diff in the ETags.
One option could be to use grep or some other method that doesn't depend on YAML or any other single format. However, a more sensible solution should prevent the ETag being overwritten on any failure