I believe that the today audio workflow is the weakest part of Kdenlive. The first step to improve it is to add an audio mixer on which to base the new workflow and possibly in the next future implement a multi channel audio support too.
We could definitely consider borrowing some code from Qtractor for the mixer:
I could try to contact Rui Nuno (qtractor's main dev) and ask him for help.
Another thing that would be most welcome is support for LV2 filters. (Although this seems to be an MLT issue) and it is in the roadmap: