I have been fascinated by this class of problem for years. In theory, if you have highly motivated, engaged, efficient people with not too many deadlines, you don't have to solve this explicitly: just let people do what they think is "important" and it'll work out. In small companies, probably like early-day Google, this informal system works out great. Once you have formal systems, you have incentives, and once you have incentives, you have to have high-level people (who are further from the actual problems being solved) make decisions about what to incent and disincent. Unsurprisingly, those decisions end up being mostly about "strategic direction" and not about day-to-day manageability, because executives don't have to do any day-to-day management. Instead, they just see the technical debt slowly build up and the teams slow down, but nobody can quite tell how it happened.