Alright, I’ve just deleted (OSMCha ) the short piece of “virtual” waterway=river that was feeding into that pipeline, and tagged the latter as waterway=pressurised (and did the same at a few similar locations around). That should separate the whole Mali Begej system from the Tisza, and let the Tisza on its merry way into the Danube. 
When the waterway map refreshes, it should show an endpoint of ~200 km rather than 20,000 km; that’s arguably correct, since the Mali Begej constitutes a separate hydraulic system, so the water of all the infeeding drains must be pumped into the Tisza.
Similarly, in @waldyrious 's examples, “virtual” waterway=rivers that feed into pipelines should be somehow eliminated…
However, as we recently discussed in Should river lines be mapped through lakes… , those “virtual” waterways that feed from rivers into pipelines are beneficial for the waterway topology analysis, so I’m uneasy with their immediate deletion. I’m inclined to just connect the pressurized ways to the main waterway, since we don’t know where’s the underwater end of that pipeline anyway (and it’s all just an approximation).