The rumors from inside Google is that there's no way to get credit/be rewarded as an employee simply for maintaining an existing solution, or even for fixing obvious breakage. Every incentive is tilted towards starting new projects, often multiple projects in the same domain directly competing with one another. There are some positives to this of course, but clearly it's being overdone.

The only way to get promoted at Google is to launch new things.

I worked at Google 2006-2010, and from where I sat, Google's biggest problems were (1) rapid erosion of corporate culture over that time frame and (2) weak project management.

A friend was managing a project on a shoestring budget. Upper management (C-suite) had reviewed the idea and green-lit it. He had been told Larry and Sergei in particular had voiced support. He got it done ahead of schedule and under budget by managing a stream of off-cycle interns. The week it was scheduled to go live, someone in middle management killed it. My friend and his team got zero credit for a job very well done, a big setback for him. He and I were working on an internal tool for datacenter management as a 20% project when I left. I asked him about it later, and our 20% project met a similar fate: enthusiastic support from management, including giving us some resources, all the way through completion, followed by cancellation shortly after completion. My friend left less than a year later.

Another friend started a modest improvement to chat as a 20% project, which later got expanded to a full-time project for several engineers. I forget the external name, but the internal code name was "Taco Town" after the SnL skit. Walkabout / Wave was a skunkworks project that used its separate repository, which was very rare at Google. People knew something mysterious was going on down in Australia, but we really had no idea what it was, other than the Google Maps guys were running it and it was named "Walkabout". When Walkabout / Wave came out of skunkworks mode shortly before external launch, the Taco Town team realized they needed to launch very soon or their project would never launch because its functionality was subsumed by Walkabout / Wave. Taco Town rushed its launch, was a bit buggy and had some scalability issues that they knew about, but expected to be able to improve shortly after launch. I think Taco Town's botched launch a few weeks before Walkabout/Wave contributed to initial confusion around Wave and some of Taco Town's problems colored perceptions of Wave's launch.

Shortly after I left, Google publicly announced they'd be putting "more wood behind fewer arrows", which was a step in the right direction.

I get it that management doesn't want to discourage engineers or stifle innovation, and they know they don't have a good handle on what will be successful and what won't, but keeping around zombie projects gives engineers false hope. The "throw mud at the wall and see what sticks" style of project management can be soul-crushing for talented junior people managing small projects, unless they're properly supported and really get proper credit for doing a very good job engineering something that fails for non-engineering reasons.