Game-modes are not "already done". They're unmaintained while the primary modes move on. This introduces bugs to the system. Additionally, some modes are unusually prone to bugs because they violate a valid assumption about gameplay. URF might create problems due to higher-than-attainable CDR, for instance.
All-4-one is insanely challenging, theoretically, because it violates an assumption that otherwise works: a champion only exists once per team. This allows optimization choices that are computationally fast that would break rapidly in all-4-one. For instance, one thing you can do is store pet-data by champion-name instead of player name. That would make it easier to access, but would snap in all-4-one, and be non-trivial to fix, at least potentially.
So, each game-mode gets a round of testing before being released again, because it might've broke since last time. What they're doing is taking this labor-time and allocating it to all-4-one development later on, instead of keeping up the schedule with other modes.
By comparison, the Project game-mode was actually relatively easy, theoretically, because it didn't violate any programming assumptions Riot might normally make. (I assume) Because of this, they mostly just had to create art assets and test gameplay. There weren't any user-predictable programming issues.
(This is all me guessing, as somebody who doesn't work at Riot, based on my experience as a programmer)