Okay, it’s halftime, so why not go play around a moment or two, eh? I have the NO side of an e-stop button connected to the FEED_HOLD connector on my Carbide Motion 2.4e controller. I fired up the Shapeoko and Carbide Motion (CM4) and ran a bunch of tests.
- Homing cycle? Feed hold pauses motion, homing cycle failed message.(Restart homing.)
- Rapid positioning tab (e.g. NW N NE…)? Feed hold pauses motion.
- Long rapid position (e.g. G0X0Y0 from across the table)? Feed hold pauses motion.
- Running a job’s G-code? Feed hold pauses motion.
- Probe cycle, Z-only? Feed hold pauses motion, cycle resumes and completes.
- Probe cycle, Z and more? Feed hold pauses motion, cycle resumes but the next X/Y goes in the wrong direction, i.e. away from the probe block. (Reset the machine.)
So, by experimentation, it seems that the using a NO e-stop button connected to the FEED_HOLD connector is a cromulent software-reliant emergency pause, except for probing (more than Z-only), which gets funky on resume. It stops all motion (unless the controller’s kicked off a robot uprising), and allows you to easily resume once you’ve moved that annoyingly obstructive clamp that you were sure had plenty of clearance. 
(I have an IoT Relay on the NC terminals of the e-stop button, so hitting the button triggers FEED_HOLD on the controller (via the NO terminals) and kills the router power (via the NC terminals), stopping stepper motors and spindle.)