I've seen highly paid engineers put on projects that had high visibility but had, in my estimation, zero percent chance of actually being deployed due the risk involved. These were often rewrites of huge subsystems to use some "new" technology, when they would've been better off spending their time simply optimizing existing code. I was right in all of these cases. The projects went no where. PRs sat un-merged for years. This was during good times though, so the people did stick around.
