This is because org-babel-load-in-session uses it like that. I think I just copied this from somewhere when I was learning how ob-babel worked. I can see now that org-babel-load-session:prolog is not doing everything it is supposed to. It doesn't load body into the session as the documentation for load-in-session says. We should fix this.

Can you elaborate on your understanding of what org-babel-load-in-session is supposed to do?

Neither micanzhang/ob-rust nor krisajenkins/ob-translate (commits referenced are those used by MELPA as of this posting) implement org-babel-load-in-session. The built-in emacs-lisp babel also appears to not implement the function.

I have to investigate when it is used. It looks to me that org-babel-load-in-session uses it to load the source block with the variables into the current session and then switches to a window with that buffer in it.

org-babel-load-in-session doesn't seem to be implemented by a lot of modes, but it seems org-babel-load-in-session is bound to C-l.

load-in-session is a built in org-babel command.

If "session" means what I think it does, then a working implementation of org-babel-load-in-session seems that it could be very useful for Prolog. In the meantime, do you think it would be prudent to remove the current, nonfunctional definition of org-babel-load-in-session?

Do you have any guarantees that the other code dealing with sessions is behaving correctly?

The org-babel-prolog-evaluate-session code should work. It did last time I worked with it. The org-babel-load-session:prolog code doesn't do what it is supposed to as it only opens the session. It doesn't actually evaluate the current block.

I see that other modes that implement session handling, like ob-R, they also have a org-babel-prep-session:MODE function. That converts the stuff like variables etc.

It should be fairly simple to refactor out the clause loading from org-babel-prolog-evaluate-session and create a new function that can be used in prep-session. Then load-session can use prep-session. The clause handeling is a little strange because of how the prolog repl behaves, but it shouldn't be too difficult.

I am also wondering if there are other commands that org-babel expects that it would be good for prolog to implement.