You always want all the accessibility features. If your button is a button, then you want the whole lot: the correct role, tabability, keyboards shortcuts, everything. There is no situation where you only want some of those features.
So if you use a div, you _always_ need to add _all_ of the features. Whereas if you use a button, you usually need to remember to add the correct "type" attribute (although if you're building a form, you might not even need that).
You also mention transparency. A button makes your design transparent: it is a standard element, and web developers should know what it does. A div is not transparent - firstly, if I'm a developer reading that code, I need to look at context clues to understand what the authors' intent was (a tabbable element with a keydown handler and a click handler could be all sorts of elements, not just a button, so now I need to inspect event handlers and see what they do), and secondly, many people do not implement this stuff correctly, so what you're usually looking at is something that looks like it might have been intended as a button, but is missing a bunch of features. Was that intentional or did the author just forget some stuff?
This isn't really a "pros and cons" type of thing. Just use a button. It's doing all the work that you'd have to do manually, but automatically. Just use it.
So if you use a div, you _always_ need to add _all_ of the features. Whereas if you use a button, you usually need to remember to add the correct "type" attribute (although if you're building a form, you might not even need that).
You also mention transparency. A button makes your design transparent: it is a standard element, and web developers should know what it does. A div is not transparent - firstly, if I'm a developer reading that code, I need to look at context clues to understand what the authors' intent was (a tabbable element with a keydown handler and a click handler could be all sorts of elements, not just a button, so now I need to inspect event handlers and see what they do), and secondly, many people do not implement this stuff correctly, so what you're usually looking at is something that looks like it might have been intended as a button, but is missing a bunch of features. Was that intentional or did the author just forget some stuff?
This isn't really a "pros and cons" type of thing. Just use a button. It's doing all the work that you'd have to do manually, but automatically. Just use it.