The current Collada exporter adds the slot index of the associated materials to each objects polylist.
example with two objects and 2 materials and different slot arrangement:

<geometry>
<mesh>
<polylist material="green1" count="256"/>
<polylist material="red2" count="256"/>
</mesh>
<mesh>
<polylist material="red1" count="256"/>
<polylist material="green2" count="256"/>
</mesh>
</geometry>

further down in the visual_scene we find:

<visual_scene>
<node>
<instance_geometry>
<bind_material>
<technique_common>
<instance_material symbol="green1" target="#green-material"/>
<instance_material symbol="red2" target="#red-material"/>
</technique_common>
</bind_material>
</instance_geometry>
</node>
<node>
<instance_geometry>
<bind_material>
<technique_common>
<instance_material symbol="red1" target="#green-material"/>
<instance_material symbol="green2" target="#red-material"/>
</technique_common>
</bind_material>
</instance_geometry>
</node>
</visual_scene>

I believe that this name scheme is not necessary. I have tested that at least blender can reimport dae files correctly when the attached patch is applied. This would make the names more consistent as there is no reason why we have a green1, green2, red1 and red2 symbol. Can someone please verify that my thought is correct and if it makes sense then apply the patch ?