With current cover caching headers, the number of requests per page is
generally lower (except for one single spike when image cache expires).
It's therefore acceptable to lower the amount of connections in the
pool, which in turn decreases the amount of needed memory.
After months of usage, I rarely paginate the list, so it's not necessary
to load 100 at a time. 20 is enough.
By matching 20 records with a 20 pool size, we should keep the repo
queue time to a minimum.
Claudio Ortolina