observed crash after reload with pending metrics

[grobian]

[2016-02-29 13:43:50] (ERR) shutting down 10.198.35.87:2003: waiting for 1 metri
cs
[2016-02-29 13:43:50] (ERR) shutting down 10.198.143.47:2003: waiting for 10 met
rics
[2016-02-29 13:43:50] (ERR) shutting down 10.198.134.33:2003: waiting for 2 metr
ics
[2016-02-29 13:43:50] (ERR) shutting down 10.198.148.25:2003: waiting for 1 metrics
[2016-02-29 13:43:50] (ERR) shutting down 10.198.148.12:2003: waiting for 2 metrics
[2016-02-29 13:43:51] (ERR) dropping 1 metrics for 10.190.145.37:2003
[2016-02-29 13:43:54] (ERR) failed to write() to 10.190.24.34:2003: uncomplete write
[2016-02-29 13:43:54] (ERR) dropping 3000000 metrics for 10.190.24.34:2003
[2016-02-29 13:43:54] (ERR) dropping 1 metrics for 10.190.139.34:2003
[2016-02-29 13:43:54] (ERR) dropping 1 metrics for 10.190.140.8:2003
[2016-02-29 13:53:25] (MSG) starting carbon-c-relay v1.8 (2016-02-23), pid=41710

relay probably crashed somwhere, no evidence in any logs though

TODO: add sigsegv handler to at least spit out a message we crashed

[grobian]

after some digging, it appears above commit fixes this crash, as it had nothing to do with reload