After upgrading Node to v4.6.0, Change-Prop's master process starts failing after initialising its workers:
```
[2016-10-11T12:20:30.739Z] INFO: changeprop/2 on scb1001: master(2) initializing 24 workers (levelPath=info/service-runner)
[2016-10-11T12:20:35.148Z] INFO: changeprop/136 on scb1001: Kafka Queue module initialised (levelPath=info/change-prop/init)
[2016-10-11T12:20:35.150Z] WARN: changeprop/136 on scb1001: listening on *:7272 (levelPath=warn/startup)
...
[2016-10-11T12:21:35.927Z] INFO: changeprop/4532 on scb1001: Kafka Queue module initialised (levelPath=info/change-prop/init)
[2016-10-11T12:21:35.930Z] WARN: changeprop/4532 on scb1001: listening on *:7272 (levelPath=warn/startup)
[2016-10-11T12:21:37.521Z] FATAL: changeprop/2 on scb1001: false == true (err.actual=false, err.expected=true, err.operator===, err.generatedMessage=true, err.levelPath=fatal/service-runner/unhandled)
AssertionError: false == true
at SharedHandle.options.process.on.process.on.SharedHandle.add (cluster.js:84:3)
at queryServer (cluster.js:488:12)
at Worker.onmessage (cluster.js:436:7)
at ChildProcess.<anonymous> (cluster.js:749:8)
at emitTwo (events.js:92:20)
at ChildProcess.emit (events.js:172:7)
at internal/child_process.js:696:12
at nextTickCallbackWith0Args (node.js:420:9)
at process._tickCallback (node.js:349:13)
```
The release notes reveal that there were some changes introduced in v4.5.0 concerning the `cluster`module, which might have caused a regression:
- https://github.com/nodejs/node/pull/6981
- https://github.com/nodejs/node/pull/7132
- https://github.com/nodejs/node/pull/6902
- https://github.com/nodejs/node/pull/6909
An initial investigation proved it impossible to reproduce the error: it would appear only on `scb1001` when the service is started via SystemD; starting it manually does not cause the error, nor does starting it on `scb2001`. Also, the previous iteration of this error was connected to UDP sockets. However, trying to disable metrics sending does not make the error disappear.