Maybe, if your cluster shuts down your streaming will be interupted.

But in your case that is probably not the issue as it seems you are not running a long running streaming query.

But what makes you think you have missing records?

Did you count the #records incoming and outgoing?