What are some characteristics of top quality research work in math?

Since you ask about top quality research, I'll say some things about the extremes. Perhaps you can extrapolate a bit from the extremes to come to an understanding.

The first kind of superlative mathematical work is one that settles an old problem that many have worked on unsuccessfully in the past.

The second sort, though it may take a while to recognize it as such, is a paper that opens an entirely new field of mathematics. Sometimes the originator may not even recognize his/her work as a fundamental advance.

So, really good math papers are those that, perhaps, approach one of these extremes in some way. An old, settled, result proved with a new technique might be interesting if the new way of proving something lets others think in a new way about other problems.

Non mathematicians often think of mathematics as a bunch of facts. Early learners in mathematics think of it as proving theorems. But before you can have a statement of a theorem you need the insight to see what might be true and provable from what is already accepted. Some of those insights turn out to be valid, others not. But it isn't about the facts, nor about the proofs of the facts, but an exploration of what might also be true and provable. If you can do that, you are doing real mathematics.