If we think about what the context menu is called then there is the argument that its contents should relate to the object that has been right-clicked. In other words, the context menu should - using this argument - only hold actions that are relevant within the "context" of that object. To say it a bit differently, the context menu should only show actions that can be applied to the object that is selected.
For instance, if you right-click on an image frame then - using the above argument - there should only be actions in the context menu that can be applied to an image frame. That's a pretty reasonable argument but there's a slight problem: What about actions that are relevant to the context but are currently either not available or not applicable?
To expand the above scenario, if you right-click on an image frame you get a list of actions that relate to that frame - all well and good so far - but you may also get some actions that are not actually applicable. For instance, there's currently an action for "Adjust Image to Frame". This action is always available, even if the action has just been applied and the image already fits the frame. Using the argument above, this action should not be available as there is no reason to fit the image to the frame again; it's a pointless action - at this moment - because it has no effect. (The same goes for "Level -> Raise" when the object is at the top of the stack. etc.)
So now we have the question of what to do with this action. Should it be greyed out - to show the user that it might become available in the future - or should it not appear in the menu until it can be applied?
There is an argument that the action should not appear because it can't be applied and having it there takes up valuable space forcing the user to mouse over it to get to other actions. There is also the counter-argument that the action should appear - but be greyed out - so that the user knows that it's a possible action but it's just not available at the moment.
The discussion about which argument is correct has gone on for decades and will probably never conclude satisfactorily to both sides but it leads me to the next point: You either clamp down on the available actions and only show those that can be applied at the moment in time the user right-clicks, or you show more actions than those which are applicable. If you show more, then how many more do you show? Who decides which are to be shown and how far do you go?
As has already been mentioned, some applications - like GIMP - offer the user all possible actions but some are greyed out and unavailable. This means that you have a very large context menu with a lot of potential actions which the user can browse through to see if they can find one they want to use. This can be very useful to casual users where they know they did something a while back that performed the action they wanted but they can't quite remember what it was called. On the other hand, it can mean that the user is shown a multitude of choices, some of which are not available and they don't know how to make them available, and that can be frustrating.
Personally, I think that the main menu should offer every possible action - with some greyed out if they are not currently available - and the context menu should only offer those actions that can be currently applied within the context of the object that has been right-clicked. (By way of example, Scribus already does this by not showing the "Adjust Image to Frame" action if an image frame has no image.) I think this is a reasonable compromise between the two arguments. People who want to see every possible action - even if they can't actually preform them at the time - can see them through the main menu and people who only want to see the actions they can currently perform can use the context menu.
To expand on that a little: There also should be no action available via the context menu that isn't also available via the main menu. To rephrase that, there should be no actions that are "hidden" in the contextual menu. There are some people who don't use the contextual menus. There could be many reasons for this: they are using a pointing device that only has one button; they have limited physical ability and cannot use more than one button; etc. These people should not be given a worse experience of using the software because they cannot access part of it.
Then there is then the argument between having a "long" context menu and having a hierarchical context menu. Again, the arguments about this have been going on for decades and there is no "correct" answer. Some people like to have everything available on the first level of the menu while others prefer a shorter menu with sub-levels. I have used some applications where a long single-level menu is good and I have used some applications where a multi-level menu is good. It really depends on the application.
I think Scribus 1.4.6 has a decent compromise between the two. For instance, I like to be able to get to "Adjust Image to Frame" straight from the root level of the menu rather than having to go through something like "Image -> Adjustments -> Adjust Image to Frame". Conversely, I prefer to use "Level -> Raise" rather than having the Raise, Lower, Raise to Top and Lower to Bottom actions all in the root level. And when you consider the "Send To Layer" actions that makes even more sense.
As for which actions should be in the context menu and where they should be in it, I don't think anyone is going to agree with anyone else on everything. For me, having "Info" as the first action isn't very useful. I hardly ever use that function and having to mouse over it every time makes the menu feel awkward. The same goes for the Undo function (I generally use CTRL+Z). I also use "Send to Layer" much more often than I use "Edit Image" or "Attributes". But that's just me, other people might use Scribus differently. I don't think there will be consensus on this.
In summary then, I think that:
* The main menu should show all possible actions - even if they're not currently available.
* The context menu should only show actions that are applicable to the object currently selected.
* The context menu should present a compromise between a single-level list and a hierarchical list.
I realise that this doesn't answer a lot of questions but I hope I've been able to explain some reasoning in a decent way that helps the discussion move forward.