There seems to be a regression (in 4.0?) in SYM_ENCRYPT and ASYM_ENCRYPT dropping messages around topology changes (and possibly bypassing retransmission?). QE tests using these protocols observe a lot of the following:

[JBossINF] [0m[31m04:45:10,048 ERROR [org.jgroups.protocols.ASYM_ENCRYPT] (thread-9,ejb,perf18) perf18: received message without encrypt header from perf21; dropping it

which is followed by numerous infinispan exceptions/timeouts, see WFLY-10464.