Why does enabling session replication trigger cascading failures in JBoss EAP 7?

Issue

• We have a JBoss EAP 7 cluster consisting of 2 instances(instance#1, instance#2) with HttpSession replication enabled.

• When instance#2 is extremely slowdown but it does not hang completely such as Full GC loop, requests for instance#1 also fail with the following message despite instance#1 being healthy. Subsequent requests with the same JSESSIONID will fail on instance#1. Requests for another JSESSIONID will not fail. The sticky session is configured appropriately in the reverse proxy.

  14:20:34,412 ERROR [io.undertow.request] (default task-2) UT005023: Exception handling request to /echo/session: org.infinispan.util.concurrent.TimeoutException: ISPN000299: Unable to acquire lock after 15 seconds for key SessionCreationMetaDataKey(e0g8ISqq1eDOgaMxMkXu_Zq081BT6Y5-pxbzUwuW) and requestor GlobalTransaction{id=9, addr=infra1, remote=false, xid=null, internalId=-1}. Lock is held by GlobalTransaction{id=8, addr=instance1, remote=false, xid=null, internalId=-1}
  

• Why does cascading failure occur when HttpSession replication is enabled?