I know this looks like the most stupid beginner question ever and I really seem to not get something absolutely fundamental about Houdinis wiring:

I want to achieve the following thing: An emitter object spawns a particle, this leading particle travels and leaves a trail of particles with the POP replication node. Now after lets say 10 frames I want another replication node to spawn a new particle at the location of the leading particle, this branching particle then flies of in another direction. All nodes write to the same particle stream “Stream_Main”
(First Screenshot)

But now of course at frame 10 the replication node “Branch” replicates not only the leading particle but also the trail particles. Makes sense. So I let the emitter write into another stream, lets call it “Stream_Leading” and the replicator for the trail writes into “Stream_Trail”.
Then I set the group field on the replicator for the branch to only work on the stream “Leading”, so only leading particles can branch of.

And of course, it is never as easy as thought. The trail particles still spawn, but no branching particles, not even in the spreadsheet view. I suspect that the replicator node for the trail somehow only lets its own particles throught, so I started experimenting with merge nodes (Screenshot 2), but now I get an unwanted copy of my leading particle. Of course, because the emitter node is connected twice to the merge node, ergo 2 particles. But why does this particle go through my replicator node and in the previous setup without merge node it does not seem to?

What do I not get about wiring? Whats my misunderstanding? Help is greatly appreciated

Okay, i made it work by trying every possible solution, but don't have any idea, HOW it works.
(See Screenshot)

How can the replicator node “Trail” replicate anything if nothing is connected to it's input?

I had the assumption that houdini works top to bottom in networks, so if nothing is at the top of a replication node it should not work.

Please, help. I have consultet countless tutorials and lession but none of them seem to explain the most basic thing about wiring and networking.

The dynamics network actually works quite differently from the geometry network. In a geometry network, the geometry data flows from one node to the next. However, the dynamics network is about attaching description data to simulation objects, and not so much about passing data from one node to the next.

The dynamics network builds up a data tree that describes the simulation setup for the substep. You can inspect this data tree from the Geometry Spreadsheet, which already looks very different from the geometry network's. The simulation is then executed according to what's described by the data tree. The documentation explains this as well:
http://www.sidefx.com/docs/houdini15.0/dyno/top10_medium [sidefx.com]

Hope this helps!

How can the replicator node “Trail” replicate anything if nothing is connected to it's input?

Thanks qsc99 this solve my problem also .
It's possible due data attaching but was not very straight forward with particles.