There is the flea, who jumps from question to question in a seemingly haphazard manner, looking for food. Or the butterfly. We admire its speed, its brilliance, and its ability to connect seemingly disconnected ideas. It's like exploring the world of possibilities by breadth-first-search.
There is the mole, slowly digging forward with obstination, going into one direction and never giving up. We admire its obstination and courage to go through even horrible calculations or case analyses. It's like exploring the world of possibilities by depth-first-search.
Then there is the rest of us. We do some kind of mixture. When we collaborate with a mole, we get frustrated by its doggedness at times, when it is clear to us that the way forward to blocked and that those heavy-handed attempts will most likely be fruitless. When we collaborate with a butterfly, we get frustrated at times, because it keeps giving up and switching directions before we've had time to see whether that previous idea had any promise.
The best collaborators are those who are in a narrow range around our own style of research: when we switch questions, they are willing to give up because they're getting tired of getting nowhere. When we keep going, they follow happily because they think there might still be something to that idea, and, although they personally would give up, are willing to give it one more try for our sake.