Well my situation is that I do not want to commit these associations to the DB for the sole purpose of keeping my database clean, the scenario I have:
Perhaps I just don't understand or misunderstand the difference between things associated and things in database, I think that the objects associated are in memory for that user and take up less resources from the server, where as if I would query the server database this would have much more impact.
I just don't understand that if I pre-sort the list before associating it, it does not use this sorting, I would think that at this point in time mendix stores the context object and all associated ID's in an array in memory as if it was the relational DB table containing the mendix ID's. Yet it does not.
I've already sent this as a feature request to Mendix, and even though it is accepted, the likelihood of it being implemented any time soon is quite unlikely.
I just wish to understand the logic behind why this doesn't work because it is clearly a limitation of the data stored in memory compared to data stored in the database. And it could be helpful to know this for other situations within Mendix.
Not sure why, but here's a workaround. Place the grid in the context of the association. Then add an xpath from the grid object to the associated entity and set equal to the %CurrentObject% Token.