It would be awesome if we could get a little more accurate controls in the collision editor. It’s really difficult to make clean colliders on objects that take up more than one tile, since we can’t line them up right.
I imagine a sub-grid with a definable (horizontal and vertical) number of divisions would be something quite convenient to optionally snap to.
Note that you can already snap to tile corners while editing colliders by holding down Ctrl.
Using the arrow keys to move around the entire object per pixel, is one - currently possible - way to achieve pixel perfect connected colliders that have shared nodes that are not on the tile’s corners.
There are “Snap to Pixel” and “Snap to Fine Grid” options already, the latter snapping to a sub-grid where the number of divisions can be configured in the preferences.
Using these snapping features it should not be hard to make sure collision objects on neighboring tiles connect cleanly. Or is there a problem that I’m missing here?
Which I found in the View menu, under snapping.
It doesn’t seem to lack any functionality, but seems too hidden for its handiness. 
^ This! I had no idea it was even in there.