A lot. A classical research question, as well as a practice I suppose, is to try to deduce the author of a particular program, whether in source or in binary. Cf.:
Code with detailed comments and include some whys, really care, and have had the nightmare of others leaving.
Although I knew one person who was notorious for just throwing their predecessors code out and re writing it. It is tough to interview to exclude that type.
Most important thing - if he likes to engineer (like most of developers regardless of experience) or solving the problem (rare kind at any seniority level).
How they approach problems in the language the code is written. Variable names are also an indicator of the level of effort they put in from my experience.
There is a world of difference between the best and worst code I wrote that has nothing to do with my personality or expertise.