Earlier quoted context omitted.
Possibly, but it's much harder to intercept and mitm specific traffic at that level. On the ISP-side, that's different: they can with high certainty say that some traffic is coming from/to a specific suspect, much like a phone surveillance. This might also apply to individual service, e.g. an email provider.
If it's in a data center security services have physical access.
From my experience in a case where a previous version of that tech has been involved (though normal LE, not intelligence), they do take all the available measures to only hit the target, it's not a shot gun approach.