Within the last hour or so, the folks on Pharo Discord#Spec (notably @Stepharo, @Esteban Lorenzano, @Steven Costiou, @stephan and anyone else I left out) fixed most of the speed problem in the NewTools debugger in Pharo 9. Its stepping action when stepping into and returning from a method (when rendering a new code pane before selecting the next evaluable) is nearly as fast as stepping within one method. The remaining latency is tiny and much more tolerable compared to what it was. The latency is longest on larger methods that need more rendering time. The cause seems to be speed of rendering with the FreeType2 library or possibly a Morphic bug (but that may have been what was fixed to create the current speed up).
The P9 debugger is workable for me. I’m closer to a port.
The only serious remaining latency problems are the already described ones following key events (insert, delete, cursor moves) and mouse clicks. The #Spec folks say that this will be fixed when the event-machinery remaining in the VM is moved to the image. Can someone estimate when this will happen? I would like to beta-test and contribute.