Ian pointed out this Git User's Survey.  Sounds like this could be useful, maybe
something to do annually.

http://www.survs.com/survey?id=M3PIVU72&channel=2WXE4BVTW8

Ganesh said that he thinks this should be an on-going survey, as in set up the
infrastructure, keep collecting results and never have a completion date [ideally]

Two things we need to do here:

1. Work out how to actually implement this.  Outsource to some company
like survs or SurveyMonkey?  Install some sort of FOSS survey solution?
 Homebrew?

2. Think about what sorts of information we actually would want to
collect from our users:

- perceptions of performance?
- commands used?
- level of awareness (eg. hashed vs darcs-2 difference?)
- platforms (issue1793)

All things being equal, I think given a choice between this and
issue1793, I'd start with latter as it's totally passive from the user
point of view and we'd collect some hard data.

Looks like Google spreadsheets make it rather easy to set up surveys. 
See Petr's http://tiny.cc/bugs-en as an example.