Several aspects of serializing/deserializing data should behave differently depending on if the source and target agent clusters are the same. To do this SSV should keep track of the source agent cluster and use that knowledge during deserialization.
For now I'm adding workarounds in MessagePort to make sure to strip out some data when source/target are for sure in different agent clusters, but those checks aren't perfect (and can be removed when SSV is fixed to take agent cluster into account).
Comment 1 by mek@chromium.org
, Jan 3 2018