I think the only time I've seen OKRs work correctly is when they were super specific, singular, and very short term. The team I've seen use them generally picks something that aligns closely with something leadership wants to see progress on, that isn't a multi-year project, and is easy to measure directly.

An example: Q1's goal was around tagging of AWS resources for billing. Our CBO came down with some new practices and automation around tagging and was having trouble getting traction in our BU with getting those tags in place for existing projects.

The work was tedious, but not difficult. Progress is directly measurable (X of Y resources in Z account are tagged correctly) and the measuring is easily automated into a dashboard. It has a massive benefit for automating Change Management and Billing/Budgeting bureaucracy.

I think OKRs have a place but I also think you just can't centralize them the way various big companies do. Even the "bottom-up" version just doesn't work because by the time the OKRs hit the C-Suite they're just mangled and incomprehensible. The OKR in my example never "escaped" past middle management. All the C-Suite knew was the CBO stopped whining about people not tagging their stuff.