I provided a fix for the failing SQL.t.  It was caused by the WRE not using InnoDB tables, which are required for running the transaction tests.

The failing i18n/label test will require rewriting the test to use a better Perl parser than the one I hacked up.  PPI is a good candidate.  I'm reluctant to drop the test altogether, since it protects us from having broken labels.

The last set of failing smoke tests looks like either a WRE or Graphics::Magick core bug.

I've been nipping away at the Session coverage.  Session::Url had one untested method which is now mainly covered.

I haven't gotten around to a whole test suite coverage run yet.  If anyone had time/tuits for that it would help a lot!