My paper was in game theory, and it took me a summer to do the work. I found (or came up with, depending on your philosophy) a class of games, proved several properties about them, introduced a stability measure and proved more things about it. The conference reviewers considered the result "important" and a starting point for future research.
I do agree that systems projects can take longer, because I worked in compilers for a year, and it can take forever to get the damn thing working (even when you know exactly what it's supposed to do). I did the entire programming and solved problems that appeared on the way, but it was my supervisor's initiative.
When we were almost done, it was running out of memory on medium size benchmarks. We added a garbage collection mechanism for our framework, that would not interfere with the rest of the compiler.
There will be a paper and a patent on that work (sometime, industry research moves slower), but it was too late for me in terms of grad school admissions. Also, sadly, my supervisor there did not have ties with those schools, so maybe that's why I sounded bitter. I didn't want to imply anyone can get good letters "just because".