I would expect the cursor in your reader program to only get a fixed set of elements, more or less the elements at the time of its creation, or maybe the elements on those pages in the Btree that were populated at the time of the cursor creation. To use the BerkeleyDB as a queue, reading off events as they are added, either use a predictable key (like an increasing number) or recreate your cursor and then skip over elements that you've already seen.