Feature branching and loss of code metadata

We use feature branching extensively as a strategy for development. In short and somewhat simplistic terms, what that allows us to do is to keep a stable trunk at all times while developers implementing complex features work on a potentially broken branch until the feature has been tested adequately. At that point the branch is merged back into trunk followed by a full integration/regression testing cycle on the trunk etc.

The problem I have is as follows: During the course of implementing said feature i开发者_如何学JAVAn the branch, there are obviously a host of commits made to the branch. Each commit has comments that hold potential educational and documentational(?) value that can sometimes be very useful. When you do a mass-scale merge of the branch back into trunk all of these comments get lost and replaced with a useless "Merging blah into trunk" type comment.

Does anyone have any solutions that can help preserve that metadata? Is this even a problem in your opinion?

P.S. I'm using SVN.