Example: an actor that collects vendors from an e commerce website, where there is no vendor listing endpoint available. Instead, it must search product listings, extract the vendor from each product, deduplicate the vendors, and then fetch additional information for each unique vendor. It may take 5 minutes to find 100 vendors, but 1 hour to find 200 vendors. As the actor runs longer, it encounters more products from vendors it has already discovered and fewer new vendors, making progress increasingly slower. The number of unique vendors that can be found is therefore not predictable in advance. In this scenario, users are happy to wait and pay for the extra 55 minutes because they are getting additional unique vendors and enriched vendor data that would otherwise not be available.