There has been a spirited debate on the public-ws-addressing list regarding the resolution to LC90.
It basically boils down to what constitutes a retransmission when a reliable messaging protocol is composed with WS-Addressing and WS-Security (or one of its offspring). The WS-RM spec basically says nothing about whether or not the wsa:MessageId (assuming one composes WS-RM with WS-Addressing) property changes from one transmission of a message to a subsequent retransmission of that same message. Nor does it say anything about whether the contents of the Timestamp element (if used) are to be changed for subsequent retransissions of a message.
The IBM implementation we used in the WS-RM+SC/T interop workshop in April was implemented such that the wsse:Security header was applied for each transmission of a message, whether it was the initial transmission or a retransmission. This is consistent with the position that has been advocated by the IBMers on the WS-Addressing WG.
Suffice to say that this issue really relates to the aspect of composability of the WS-* specs and hence should be addressed in a profile. Fortunately, we just happen to have such a profile. In the next revision of the IBM Basic B2B Profile, there will likely need to be a new requirement requiring that when composed with WS-Addressing and WS-Security, that the Timestamp element carried in the wsse:Security header MUST be replaced with a value equal to the current time (relative to the RM Source) on every transmission of a message, whether it be the initial attempt or a retransmission triggered by lack of receipt of a SequenceAcknowledgement for the message (or something to that effect).