God knows I suffer from that. The beginning of something new is always awesome but as time progresses and the "new" feeling vanishes I find it harder and harder to keep working on it.
They don't want to waste engineering talent on a product that doesn't even register on their bottom line, and shows no promise of explosive growth. But on the other hand, they want to build the next big thing in-house, rather than having to acquire it for billions or letting it grow into a competitor.
It's really tough predicting what the next big thing is, so they throw spaghetti at the wall.
From talking to PMs at Google, they have exactly this problem. Engineers are free to pick and choose what they work on.As soon as a product is released, hits "1.0", and the engineers have claimed whatever kudos they can, they move on to the "next big thing", leaving the PM alone with the product. That leads to situations where PMs can't even get basics like localizations to specifics countries / languages.
If you think through the logic on 20% time and deliberately developing as many different crazy projects as possible in the hope that a few of them pan out and hit the big time, this is the inevitable outcome.