For any benefit the auto-router adds, it usually ends up costing a project later...
In general, impedance controlled UHF design work with domain specific simulation tools is the modern workflow. Thus, one manually does the critical tracks first, island-pours, and finally power interconnects.
KiCad is slightly better than nothing for layout, and trying to make it into yet another half-baked simulation tool seemed silly. =3
I wish the Kicad team had sunk all the time they put into simulation support (which I rarely use, in or out of kicad) into a proper constraint system instead.
If I need to sim, I’ll just use LTspice or TI TINA - I.e. the simulation ecosystem that includes the part model I need, with no wrangling to get that spice model to play nicely in some other spice toolchain.
For learning, the current LTSpice models and QUCS VNA profiled device imported simulation data are reasonable tools as well.
I used KiCad all the time for personal projects, and it has recently already addressed the major issues people found annoying:
1. unstable library version dependency management
2. complex design collaboration over traditional versioning tools like git (still an upstream feature last I checked, but should be available soon.)
3. freezing the specific KiCad and library path structure versions across a project, as assuming every design has a single workstation workflow is a contradictory goal
KiCad is better than most current EDA software, but that is mostly a sad reflection on the traditional tools available. =3
In general, impedance controlled UHF design work with domain specific simulation tools is the modern workflow. Thus, one manually does the critical tracks first, island-pours, and finally power interconnects.
KiCad is slightly better than nothing for layout, and trying to make it into yet another half-baked simulation tool seemed silly. =3