Deadline Propagation
Propagation of a request's remaining time budget across downstream calls so that independently chosen timeouts do not violate the end-to-end limit.
If a request must finish within 500 ms, giving a downstream service another independent 500 ms does not preserve the original bound. Elapsed time has to be deducted and only the remaining budget should travel with the next call. Deadline propagation is therefore different from assigning the same timeout everywhere.
Retries consume that budget as well. Starting another attempt when the remaining time cannot accommodate it increases load without improving completion probability. This is the same boundary examined in safe retry design in critical systems.