I don't get it. It seems like they're doing largely what they said they would.

They wanted to push a feature, and they said they would if they didn't see any major regressions. Then they did see a major regression, so they pulled the feature.

Exact version numbers, timelines, and builds are pretty irrelevant to that process. Or are you actually saying you would prefer they had just left their product broken for a significant portion of users, just to keep aligned with the version numbers they mentioned in a blog post?