I had not time to try this out yet and my brain might be missing something after a long day of work. But this should work already if project(':modules:jars:onepad_common') is execute before the current project. What happens if you execute the above?
BTW: In trunk we have now a reasonable naming schema for archive tasks 
The example in the user documentation myZipTask.merge('path1/otherArchive1.zip', 'path2/otherArchive.tar.gz') passes the arguments in a list context (which looks like a 2 argument method to a traditional programmer) but not as a list object even though the merge method takes a single list as an argument. This may be auto-magically handled by groovy (I don't know the nuances of it) but I was able to run the merge method correctly if I wrapped the single argument as a list:
task(this.project.name + '_jar').merge([ commonJar.getArchivePath().getCanonicalPath()])
{ exclude( 'META-INF/**') }}
Maybe the documentation should explicitly document list arguments where they are appropriate to avoid confusion for us groovy-ignorant gradle builders.
Keep up the good work!!!
The merge method takes a Java 5 varargs argument merge(Object... paths). We then do two things with the paths array. First we check if the last argument is a closure. If so we strip it off and apply it afterwards. Then we flatten the paths array (therefore your example above works).
In your case it would be easier to write:
{{task(this.project.name + '_jar').merge(commonJar.getArchivePath().getCanonicalPath()) { exclude( 'META-INF/**') }}
If this does not work, that would be a bug.
But you can also do: merge(someListWithPaths, aPath, anotherPath, anotherListWithPaths)
There has been a bug in the flatten behavior which occured under some circumstances. This is fixed in trunk.