I've heard from the industry that the issue with GSA was that it just wasn't that good. It tried to adapt Google's core search into a box, but Google's algorithm is only great because of scale. The box doesn't have as much of an opportunity to learn from user interactions and thus fails to meet expectations.
The GSA "just worked". Also having the name recognition for public search did a lot to drive its enterprise popularity. It didn't have a lot of bells/whistles, but it addressed the commodity search problem well.
Agreed relevancy is a problem. PageRank works well for public content but internal search has plenty of relevancy problems. Having little control over this certainly hurt.
Managing on-prem hardware/appliances is a difficult business. I don't fault them for moving to a more scaleable model.