Same applies really, been years since I personally clicked "render" without it being a redirected shell command doing the actual render...
Hope that makes sense, as in even if you are rendering just 1 frame full bounding box etc etc etc Id have thought you'd still be running a single command line render to get that frame
The two methods don't compare, one is clicking things and then running into scenarios as you describe :) the other is scalable for things like this, by performing your renders always command line no matter how small you get them any modification is a small addition to your pipe rather than a new concept is what I'm trying to get at I guess.
There is a way to do this in the UI of course, would throw you the example but am travelling up near the arctic circle and don't Maya with me.
anyone else feel like writing either a snippet which either replaces the default render button with a default render shell command picking up from the render globals so the -reg flag is passable to it? ( isn't the option already there in the render globals? , called region of interest it something like that )
Or go the complex way about it and build a container, split it into cubes and control your segments in 3D space, you get some insanely fast simulations and exponential renders when they start affecting each other but I'm not really sure you want to go down that route.
If all you are after is just a "click and show me this part" , rather than splitnStich a few terabytes of data into a complete image the answer is within the render globals, the same as where you found the Arnold ones.