memcached / ep-engine crashes when client doing UPR restart

Upr server fails when client restarts UPR streams with the same name. The sequence of operation on the client side,

• client acquires a connection say port number 62138 for UPR feeds.
• client then acquires another connection say port number 62139 for memcached.Set() operations.
• client gets the mutations via UPR feed (via port 62138)
• client then closes the UPR connection (port 62138)

• client then re-uses the open connection (port number 62139) from connection pool to restart UPR feed. I am using the same UPR_OPEN `name` as before.
• client then acquires a new connection say port number 62140 for memcached.Set() operations.