Currently the behavior is to poll the servers in a loop, waiting until the observe timeout to fail the response. In the case of a configurable user application where e.g. a user may wish to sometimes disable persistence/replication, a use case may be to call the ExecuteStore function with Zero, Zero as arguments for persist and replicate.
In this case, the internal observe implementation will wait a bit of time and then fail the command. Neither is desired behavior, as persistence is not expected and should be no-ops.
[ There are probably other bugs lurking in this behavior which haven't been discovered yet.. ]