Proofing The Proof Corrections?

OK...this has happened to me more than once so I would like to comment and hopefully receive some feedback. You send in your final revisions for your manuscript and it is accepted by a journal. Sometime later you receive the proofs with the layout for final publication and a request for you to check it over and correct any final typos and/or problems with the layout. You submit your proposed changes within the time alloted and sometime later the manuscript is finally published. Upon reading through the published manuscript you see where some but not all of the changes you requested in the proofs were made, and thus there may still be some annoying errors (typos and/or layout) that you thought were going to be fixed. I am not talking about major changes being requested (which are costly at the proof stage and thus may not be approved by the editors). Instead this is about small corrections that for some reason were not made.

What do you do next? Of course you can contact the editors and point out were requested corrections were not made, but what does this approach get you? An apology and possibly an erratum in the next issue, but does not fix the actual publication (which is now a permanent record). Maybe it is possible to actually proof the proof corrections? I'm not aware of any journal that does this, but I may start asking.

I guess that hypothetically through the writing, review, and rewriting stages all mistakes should have been eliminated, however, anyone who publishes knows that this is not the case. In fact, the proof stage is one of the most important parts of the whole process because it represents the first time the manuscript has been out of the author's control and as I said earlier it represents the permanent record of your work. Wouldn't you like just one last look before it finally does print? I actually had one manuscript a few years back where we did not even get proofs! The final result was not good and we actually felt obliged to add a disclaimer to the reprints and PDFs that we distributed.

Has anyone else had a problem with this? How common is this type of error?