From discussion today: For simple analyzers, this could be a script that goes through a sequence of CL numbers. For each CL, it could get files for that change and run an analyzer either through tricium or not, either on remote swarming or locally. For example, it might:

  Make a bare git repo clone, e.g. of chromium/src
  Fetch the analyzer from CIPD (or pass a local analyzer binary path)
  For each CL in some sequence of CL numbers:
    Clone the bare repo to a temp directory
    Check out the files in the CL (like GitFileIsolator)
    Run the analyzer on those files
    Copy the results somewhere if there are any results
    Clean up the temp repo
  Print a summary of the results

Also, depending on the use-case, running the analyzer on HEAD in the repo will likely give a pretty good indication of roughly how often the analyzer will have results.

Another point here: it's possible to run Tricium on arbitrary git refs for arbitrary lists of files, so we may just want to request analysis for some git commit for some large list of files.

It may be convenient to add a CLI (bug 718396) for this.