Ah, Issue18604! Thanks for the reminder (and for the patch), Serhly! Focusing on these more recent occurrences of segfaults, I hadn't been thinking about those earlier ones.  Now, after digging into Tk and identifying the problem there, it's clear that those earlier segfaults reported in Issue18604 were a result of the same Tk bug, just caused slightly differently from the tkinter tests.  The #18604 changes also caused a Tcl interpreter instance to be created and destroyed without actually causing Tk to do enough to cause the "open application" Apple event to be fired and received. _is_gui_available is the right place and adding the additional Tk activity there does it at the right time.  And it appears that root.update() is sufficient, as you suggested.  I want to do some additional testing but, assuming no problems are found, I'll plan to apply the simplified patch.