From a clean install of Ubuntu Karmic (x86_64) (rather than in-place upgrade from Jaunty), and mounting the existing /home/ volume into it (so the user settings are all inherited) I've checked and this bug isn't affecting this PC.

The only difference I can see with regard to Compiz settings is I have custom settings (not 'Normal' or 'Extra'' visual effects) because I have the 3D cube configured on two X screens.

It looks as if the problem is in the source package generation and the patch itself.

The patch to fix this is "debian/patches/012_default_keybinding.patch" which patches "src/metacity.schemas.in" which is generated by automake (from "src/metacity.schemas.in.in"). These lead to 'make' creating "src/metacity.schemas".

However, when the package is built 'make' regenerates "src/metacity.schemas.in" thus wiping out the changes made by the patch:

------
$ fakeroot debian/rules build
...
Trying patch debian/patches/012_default_keybinding.patch at level 1 ... success.
...
Generating keybinding schemas... ./metacity.schemas.in.in
./schema_bindings ./metacity.schemas.in.in ./metacity.schemas.in
LC_ALL=C /usr/bin/intltool-merge -s -u -c ../po/.intltool-merge-cache ../po metacity.schemas.in metacity.schemas
Generating and caching the translation database
Merging translations into metacity.schemas.
...
------

This is obvious if you try to do a package clean:

------
$ fakeroot debian/rules clean
...
Trying reverse patch debian/patches/012_default_keybinding.patch at level 1 ... 0 ... 2 ... failure.
make: *** [reverse-patches] Error 1
------

So the obvious fix would be for the patch to be applied to "src/metacity.schemas.in.in".

However, looking at the contents of the patch it isn't clear to me *why* that should work since it affects the "/apps/metacity/window_keybindings/move_to_workspace_12" key, not "/apps/metacity/global_keybindings/switch_windows_backward".

Also, the patch is described as a fix for this bug in comment #15 and a changelog entry is shown there which is in the Hardy package changelog; however that entry *does not* appear in the current source package's changelog (or any since Hardy), instead the patch itself is referenced but no specific mention of this bug reference:

------
metacity (1:2.23.21-0ubuntu1) intrepid; urgency=low
...
  * debian/patches/012_default_keybinding.patch:
    Fix default keybinding for shift-alt-tab 

 -- Pedro Fragoso <email address hidden>  Tue, 03 Jun 2008 18:36:06 +0100
------

Inspecting the contents of the patches for Hardy (created my Michael) and the patch that is in Intrepid and later (from Pedro's changelog report) the difference is obvious - they are being applied to different key-bindings:

------
$ diff -Nu metacity-2.22.0/debian/patches/012_default_keybinding.patch metacity-2.28.0/debian/patches/012_default_keybinding.patch
...
+@@ -1348,7 +1348,7 @@
+       <applyto>/apps/metacity/window_keybindings/move_to_workspace_12</applyto>
        <owner>metacity</owner>
        <type>string</type>
 -      <default>disabled</default>
 +      <default>&lt;Shift&gt;&lt;Alt&gt;Tab</default>
        <locale name="C">
-          <short>Move focus backwards between windows using popup display</short>
-          <long>
+         <short>Move window to workspace 12</short>
+         <long>The format looks like &quot;&lt;Control&gt;a&quot; or &quot;&lt;Shift&gt;&lt;Alt&gt;F1&quot;.
------

So the issue seems to be twofold:

1. The patches are inconsistent, conflicting, or accidentally misformed during a source package refresh
2. The patch should apply to "metacity.schemas.in.in" if the package-build is going to cause "metacity.schemas.in" to be regenerated after source package patching has been applied.