Bert Massop

I'm not convinced that this would actually work. I'd be glad to be proven wrong by simulations, though. > Local requests are randomly distributed across the keyspace. But our peers...

Apart from the style issues mentioned above, this looks good to me.

Except for some formatting issues this looks reasonable.

> On a thread-constrained node it saves up to 8 threads _when some types of request are not running locally_. That practically amounts to "never" on a connected node. I...

The benefits of this pull request to the project as a whole void my previous objections to having this merged. There are some whitespace issues with this changeset. As soon...

This looks good, except for formatting. I agree with @ArneBab that this is conceptually sound, but haven't looked at your alternative #491 yet.

If `_matched` posed a problem, `_message` is problematic as well. `getMessage` should then also be synchronized on `this`. (and so should effectively all users of all other member variables be...

This change set is duplicating implementation logic of a multi-value map all over the place. I suggest to move that logic to a separate class (for it to be reusable),...

My usual stance would be "you touch it, you fix it", but I think we can agree to disagree on this one. I can live with the proposed changes from...

> Also more generally our encryption setting doesn't necessarily match what's on disk, so we have to try both. We expect that our encryption setting matches what's on disk. Trying...