This issue was caused by timeouts between Artifactory and Access due to intermittent load on the access server. This was also compounded by a separate issue where the server did not recover from those timeouts as it should have.
During the incident, we made configuration changes that helped to mitigate the impact of those problems. We also tuned our alerting so that our teams could respond immediately when proactive hallmarks for this issue arose.
As a long-term solution, we have created a patch that allows Artifactory to recover automatically from those timeouts with no impact to the service. In addition to the patch, we've added additional logging to assist in future debugging should similar errors arise.