Exactly. Why not?Think "gravity waves", emitting from let's say a black hole.
It's an uneducated hypothesis, but why not?
The problem is that you can get it to work with galaxy rotation curves, but it's going to do bad things for nucleosynthesis and early universe and you run into some obvious problems with pulsar timing experiments. If there are graviton-graviton interactions then you ought to see "weird things" happen at strong field strengths.It was peer reviewed and published in a good journal. But who cares? It didn't get even 1 citation. (the paper mentioned by INSPIRE as citing it, in fact, does not cite it)