We are at times having some difficulty achieving a 'successful' replication.
Our Operations personnel have reported the following behavior (which I cannot
attest to): A master has a set of segment files (let's say 25). A slave then
polls the master, get the list of segment files that differ and begins to
download them. Sometime during the download, the master combines two or more
of the files that the slave is going to download and when the slave attempts
the download it fails. We're aware that a subsequent attempt usually yields
success, but I'm curious as to whether there are any configuration settings
that can help mitigate this circumstance.
