The exact same thing as if they grab the ground at their feet when being pushed away on flat ground: nothing, they still end up 10 feet away from their starting location because that's what the effect does. Put another way, if grabbing something to prevent movement is ineffective on flat ground, then it is equally ineffective on a cliff.
Perhaps there should have been a saving throw to resist the movement baked into the mechanic. I can certainly understand the arguement there. Personally I think there should have been one, though I can understand not including it for the sake of speeding up game play. Adding an additional saving throw for the DM to resolve for every hit a warlock makes with his normal attack cantrip has the potential to slow down the game significantly. It's far faster to link every hit with the forced movement, and if once in a while that throws someone from a cliff so far that they can't grab it, so be it. As a DM I'm glad that I don't have to deal with all of those extra dice, and the PCs taking a monster out of the fight instantly helps cut down on what I need to deal with. If I'm truly bothered by it I can just ramp up the pressure later.
From what I've seen, this ability is only really useful on rare occassion. Unless you're throwing someone into something dangerous (fire, Spike Growth, off a cliff, etc.) or playing movement speed games, it has little effect. To arbitrarily add an additional saving throw to resist it working is needlessly punitive. I don't give monsters a death save to resist being killed when the fighter brings them to 0HP. The monk beat the monster's AC? Great, it suffers the effects of a successful attack (1d6+4 damage and dies if that brings it to 0, Con save vs. Stunning Strike, unable to take reactions because of Flurry of Blows). The Warlock beat the monster's AC? Great, it suffers the effects of Eldritch Blast (1d10+3 damage and dies if that brings it to 0, pushed ten feet away).
