Not entirely clear if this is supported with Redis directly, but ideally a task that is still healthy and alive should be able to periodically extend its claim, which would allow folks to use the a lower redelivery timeout for more timely processing, but still have good reliability.