Google's promotion process biases very heavily towards "impact".

The problem is that maintaining existing projects generally isn't considered to be a demonstration of impact (or at least, it's a MUCH harder case to make). It's much easier to make the case that greenfield projects demonstrate impact.

Additionally, engineers have a good amount of power over what they get to work on: managers generally have to convince them to join a team. It's nice in concept, but lacks a steady influx of unwitting new engineers to help support all the not-so-sexy projects (like many other companies).

…so, you tend to see a pattern of engineers building a new thing, launching it, and then moving on to the next project. Leaving a trail of poorly supported projects in their wake.