On sites with heavy traffic, the recommendation extension can be quite a performance killer.

To make calculation and display of recommended items faster, we can introduce a cache which will avoid executing the rendering template for all the items which get recommended.
This cache will be shared by all users with same permissions which get recommended the same item set (as an alternative, the developer can currently add cache-blocks inside the template used to render the recommended items).
The cache will be configurable with a ttl setting