Answer here (courtesy of Kaiser Fung).
Solving a question with 10 minutes discussion and a Sign-test. They put my name on the paper.
The things I did that (positively) impressed researchers seemed to be things that suggested a change of question, focus or design (in which I would include using different data) rather than doing things like a frailty survival analysis before most statisticians had.
There is though, likely little overlap between what impresses researchers versus other statisticians which may explain why there is very little published in statistical journal about how to apply statistics more heroically.
Mail (will not be published)