The incident is now solved.
A cache database managing our feature flags has been failed over, without the feature flag service detecting it. It caused the feature flags service to become inconsistant, which produced application misbehaviors in the visitor livechat application. Everything went back to normal after detecting the source of the misbehavior and restarting the feature flags service.
Thanks a lot for your patience.
Apr 18, 17:56 CEST
There was an incident impacting conversation delivery between 3:35PM to 3:43PM UTC+2.
Service is restored and we are monitoring .
As a consequence, agents may have received more conversations than they should.
Apr 18, 16:30 CEST