[users at bb.net] Improper coalescing of try builds from different branches with gitlab?

[Dan Kegel]

With a recent buildbot (facc540446191a7faae3e9cb85b2248bc7ce4bc9 from July 12)
plus my GitLab stuff, I saw something funny happen this morning.

I'm building using SingleBranchScheduler (as in the gitlab sample),
with a tree stable timer.  Two different merge requests came in
during the tree stable interval, and although two changes were
injected, and two builds were started, they both thought they
were the latter of the two changes.

Is this a textbook case for using AnyBranchScheduler instead?

BTW the only reason I'm using a treeStableTimer is to suppress
building on each commit inside a merge request, so I could probably
reduce it to 1 second.  I may be abusing it, dunno.