The OAuth 2 lead author resigned from the project citing numerous issues they saw with the specification. I don’t know exactly why Valve ended OAuth 2 but it certainly wasn’t looking good for OAuth 2 at that point.
Steam keys don’t have any real need to be hidden from the owner so in most contexts the web based activation is perfectly fine. It just doesn’t work well in situations where groups of keys are given away.
One other thing that’s important in all of this: costs & benefits. Was maintaining something similar to the older activation technique going to increase Valve’s revenue by an appreciable margin? Maybe they didn’t think so.