I am concerned that the current handling of copyright problems on wikipedia may be insufficient. As it stands, after detection the offending text is completely removed.
Unfortunately, if there has been a long time span since the insertion of the infringing text there may have been a substantial number of valuable contributions to the article. With the way that most content grows organically over time, it may be very difficult to say if the new text would have been created without the infringing text with any certainty.
In the United States the recent tendency in court appears to be to favor the most expansive definition of a derived work possible. Because of this, I suspect that it would be likely that at least some of the contributions made to an article after the insertion of infringing text would be found by a US court to be derived, thus placing their ownership in question. This interpretation of derived isn't necessitated by current international treaty, and would likely be different (and possibly more sane) in other locations, but I suspect that US legality is a substantial concern.
Determining if a piece of text is derived from another, at least in the over broad sense favored by US courts, is an intractable problem, but the policy could do a better job of avoiding these concerns. Reverting to the point where a substantial amount of infringing text was added, and deleting *all* modifications after that point would be much more certain to avoid impinging on the intellectual property rights of others.
The cost of destroyed improvements might be mitigated by the benefit of creating a greater incentive for frequently contributors to quickly catch and remove violating text.
Of course, none of this is legal advice...