People hear "unlimited requests" and picture a warehouse of tickets. Here is how it actually runs on poketdev.com: every client gets one active request at a time. You can queue as many as you want, but we build them one by one, and you approve each one before the next starts.
The queue lives on a Trello board the client can see. Cards move from Requests to Active to Approved. When a card lands in Approved, the code is already sitting in the client's private GitHub repo. Nothing gets handed over as a zip file or a "demo on my laptop".
The reason I refuse to juggle five half-finished requests per client is simple: split attention is where timelines go to die. One active request means the answer to "where is my thing" is never "in three branches somewhere". It is either in review, or it is done.
It also keeps turnaround claims testable. Ours is 72 hours for a typical request, and a single-item queue is the only way a number like that means anything. If a request is bigger than 72 hours of work, it gets split on the board, in the open, before anyone wonders what happened to the week.
The setup is Mon-Fri, async-first, zero meetings. Onboarding to the board takes about a day. Cancel anytime and keep the code; the repo was yours the whole time.
If you want the audio version of these notes, that is the podcast, The Flat Rate Dev.
Next issue: pricing a flat rate when you cannot see the future. The real numbers are in poketdev.com if you want them now.