Redundant restarts

[widhalmt]

We have handlers restarting services and thus creating lag that can lead to timeouts in our checks.

I guess, we can reduce the count of restarts, especially for Elasticsearch and speed up the checks dramatically.

One issue I found so far is that in #137 we agreed on not restarting the cluster after the change, but we still notify the handler.

I'll look into the roles and try to remove any redundant restart.

[widhalmt]

This might also be connected to #252 .