PHYData DBA (11/12/2013)
I got it right not because I actually knew the answer but because I figured that the data (meaning common) pages would be fixed by automatic page repair while more high-end issues would not. The logic being that the more you need it the more likely it wont happen.
LOL... That is cute but I believe the actual logic is it can only repair the pages to do with the mirrored Data itself and not the copy of the mirrored data.
I.E. - I can request a page that is in the data pages of the mirrored database source, but not anything that is unique to the partner database created to hold the mirrored database.
I guess I agree with PHYData; some pages contain information about physical positioning (ie information about the the physical copy of the data), and since the physical to logical mapping won't be the same for both units those pages won't be recoverable. Other pages contain only logical positioning information, so are the same for all copies of the data: no reason they shouldn't be recoverable. It is of course a natural consequence of Murphy's law that the pages whose loss do the greatest damage are the ones which are not recoverable, but Murphy's law applies to all processes, including the process of making deductions from Murphy's law, and thus jvogel's reasoning doesn't wash.