this exact thing is an implicit fix we have long applied internally for all users in our pants wrapper script. the problem is well understood but we’ve worked around it for the two known cases (which both have upstream
python.org bugs - this scproxy one as well as an sqlite one, which we solve by static linking sqlite vs using the OSX system provided libdispatch tainted one). so 2 patches for observed issues now vs a long term strategy to kill the forking. it’s true there may be other cases - but we have not witnessed them at scale, so (at least I) feel fine with it w/ the bigger picture in focus.