Reporting this as a defect, although I'm not entirely sure what a good
solution is for it just yet.

Take the following scenario:

JobA polls SCM every minute, or every 5 minutes (it takes 30, so it doesn't
matter). DevA and DevB both commit and push changes which queue the job within 5
minutes of each other (commits #1 and #2 respectively).

When the first build completes (with just #1) the "Changes" will be just rev #1.
When the second build finishes, the "Changes" will be rev #1, rev #2. In effect,
you will have changes listed for the build, which will not be unique to that
particular build, but in fact were introduced and processed by a different build
number.

I believe this is a problem, particularly when hunting down test failures to
find the commits that they regressed in, since they will no longer be distinctly
associated with a given "failed" build.

I'm not sure on a resolution for this, but I thought it merited discussion