[Codex desktop] Per-project/session priority scheduling, independent of model and reasoning effort

What variant of Codex are you using?

Codex desktop app. I use Windows, but this is a general, cross-platform feature request.

What feature would you like to see?

Add a Priority control to projects and individual chats/sessions so users can decide how available usage capacity is distributed between concurrent work, without changing the model, reasoning effort, or quality settings.

Example

I have Codex working on two projects simultaneously. Both matter and should receive the same high-quality reasoning, but my main project should get a larger share of the available capacity while the secondary project progresses more slowly.

For example, I could assign the main project a weight of 3 and the secondary project a weight of 1, targeting roughly a 75% / 25% share of the relevant shared capacity while both have work ready to run. This is a relative allocation preference, not a request to charge three times as much for the same work or a guarantee of three-times-faster completion.

The analogy is torrent download priorities: allocate more bandwidth to an important download without reducing the integrity or completeness of the others.

Suggested interaction and behavior

• Right-click a project or chat/session and choose Priority → High / Normal / Low, with optional custom weights or percentages.
• Let chats inherit their project's priority, with a per-chat override. Make the selected priority visible and adjustable while work is running.
• Give higher-priority work a greater share of available execution/usage capacity. Pace or queue lower-priority work at safe boundaries rather than cutting its reasoning short, downgrading its model, or skipping validation.
• Allow other runnable work to use unused capacity when a higher-priority project is idle or waiting for user input. Keep all execution within the account's existing limits.

Important distinction

This is priority scheduling across concurrent projects/sessions, not a request for Fast mode, a different model, lower reasoning effort, or merely sorting projects in the sidebar.

The lower-priority project should wait more, not think less. Each session's selected model and reasoning settings should remain unchanged; priority should govern when/how often work runs, not how carefully it is performed.

Additional information

This would let me keep secondary projects moving without having to give them the same share of my available usage as my main project, manually pause and resume them, or compromise their quality. The request is for control over allocation and pacing, not additional quota.

The issue names projects and chats/sessions as the scope, but mentions no files, tests, or entry points. Begin by tracing how concurrent work and account limits are currently managed; done means users can set inherited or overridden priorities that pace runnable work without changing model or reasoning settings.