Forte is a good example for this. Coming into the season, no one knew what type of player he would be or what type of impact he would make. By updating the ratings, you are improving the accuracy of the league and the players and are thus making the rosters more realistic. And despite your concern that you want to the players to evolve to that point naturally, this is unrealistic in some cases. For example, perhaps Forte has a low potential rating in the game(I don't actually know what his pot. rating is). This puts a ceiling on the most he can improve within the structure of the game. And on the flipside, Tavaris Jackson's potential rating might be too high, etc. The game isn't completely open for anyone to turn into a star. So, the updates can provide this accuracy. And, this would create a new experience with every new roster update.
Another example would be Marques Colston a few years ago or even Tom Brady. There were no expectations for either of them when they came into the league but, with the current rosters, this would be impossible to replicate. If Matt Flynn(whom the Packers seem to think is better than Brain Brohm) came in after an Aaron Rodgers injury and turned out to be the next Tom Brady in real life, wouldn't you want that in the game as opposed to him being a low rated QB with a low potential rating? The only late round QB from this past draft that I've heard about doing anything for anyone is Andre Woodson. So, that shuts the door on all the other players trying to prove themselves.
Personally, the one rating I'm most in favor of consistently updating is the potential rating because I agree that the players should have to grow into that through the game. Because if Forte's potential is an 80, then that just won't do. And yeah, I'm for only updating the ratings. We can decide who to start and so can the CPU while in game.
Oh, and for the people voting for Josh & Co. to do nothing, no one's forcing you to download the roster update.