Comment 1 for bug 2042943

- I looked into the logs of those failures - they are not related to that change
- there are indeed bad messages written by the keyring lib from time to time, but they are written to the stderr what should not influence output parsing
- in case of every failure I found there were other issues (i.e. from keystone)
   - https://storage.gra.cloud.ovh.net/v1/AUTH_dcaab5e32b234d56b626f72581e3644c/zuul_opendev_logs_23d/899157/1/check/cinder-tempest-lvm-multibackend/23d8799/controller/logs/screen-keystone.txt
   - https://storage.gra.cloud.ovh.net/v1/AUTH_dcaab5e32b234d56b626f72581e3644c/zuul_opendev_logs_d2a/periodic/opendev.org/openstack/neutron/master/devstack-enforce-scope/d2a5e4a/controller/logs/screen-keystone.txt
   - https://6c40da3fb44111d52eec-5a0483cb68bfeeccc28b5042e69d7a01.ssl.cf1.rackcdn.com/periodic/opendev.org/openstack/neutron/master/neutron-ovs-tempest-with-os-ken-master/c3aaeb6/controller/logs/screen-keystone.txt
- opensearch query is a bit misleading. It selects occurences when certain string is present in the log and the build status is failure (independent whether in this job or not)

So far I have not seen a certain proof that it token caching change caused those failures