The kapacitor can't alert after the task was reloaded.

I have 10 nginx servers that are using kapacitor for alerting. The `TICKscript` as follow:

```
var data = stream
|from()
.measurement('procstat')
.where("process_name" == 'nginx')
.groupBy('host')

var alert = data
|deadman(0.0, 30s)

alert
.log('/tmp/deadman.log')
```

The kapacitor can normally alert after a nginx was dead. But It can't alert after a nginx was dead while the task was reloaded. It's so weird.

If I really want to remove a nginx server in another case, what should I do?

Start with the TICKscript showing procstat, groupBy('host'), and deadman behavior, then reproduce the alerting failure when the task is reloaded while nginx is down. Trace the task reload and deadman state handling; done means alerts resume correctly after reload and the supported way to remove a nginx server is clear.

Written by the indexing model from the issue text.