Migration test for glib2.0 unexpectedly using dbusada/unstable source code

[Simon McVittie <smcv@debian.org>]

While monitoring test results for glib2.0_unstable via the
package tracking system, I noticed that dbusada is reported as having
an autopkgtest regression with the new GLib. However, the logs seem weird:

> autopkgtest [22:14:19]: host ci-worker01; command line: /usr/bin/autopkgtest --no-built-binaries '--setup-commands=echo '"'"'dbusada testing/amd64'"'"' > /var/tmp/debci.pkg' --user debci --apt-upgrade --add-apt-release=unstable --pin-packages=unstable=src:glib2.0 --output-dir /tmp/tmp.i2Nvdf5qy1/autopkgtest-incoming/testing/amd64/d/dbusada/4386177 dbusada -- lxc --sudo --name ci-057-5af08ed6 autopkgtest-testing-amd64
> ...
> autopkgtest [22:14:41]: @@@@@@@@@@@@@@@@@@@@ apt-source dbusada
> Get:1 http://deb.debian.org/debian unstable/main dbusada 0.5.0-3 (dsc) [2,172 B]

To check whether migrating glib2.0 from unstable to testing would cause
a regression, I would have expected that we'd want to be running the
tests from dbusada/testing (which is version 0.4.x), with all packages
taken from testing, except glib2.0 (and its dependencies if they are not
satisfiable in testing) from unstable.

However, what we're actually getting is an attempt to run the tests from
dbusada/unstable (version 0.5.x). dbusada in unstable is currently broken
(its build-dependencies are unsatisfied on all architectures - I've
reported a bug), hence the regression.

Is this a bug in the migration infrastructure, or a bug in debci, or are
my expectations wrong?

Thanks,
    smcv