Function wait_all 

pub async fn wait_all(
    processes: &mut [&mut RunningProcess],
) -> Result<Vec<Outcome>>

Wait for all of several running processes to exit, returning their Outcomes in the same order as processes. The processes are only borrowed and stay usable afterwards (the exit status tokio caches remains re-readable).

Same two non-features as wait_any: no per-process timeout (bound the whole batch with tokio::time::timeout) and no output pumping (a contender that fills its stdout/stderr pipe blocks forever — drain chatty children first). Unlike wait_any, an empty slice resolves immediately to an empty Vec: collecting zero outcomes is well-defined, where racing none is not.

If a contender fails to reap (an OS I/O error), that Err is returned and the remaining processes stay waitable (cancel-safe). A contender’s Error::Cancelled (cancelled run) or Error::Stdin (a non-broken-pipe stdin-source failure on its otherwise-successful exit) likewise short-circuits the join — like the bulk verbs, these surface as an Err, not an Outcome.

Errors

A contender’s Error::Io (a failed reap), Error::Cancelled (a cancelled run), or Error::Stdin (a non-broken-pipe stdin-source failure on its otherwise-successful exit) short-circuits the join; the remaining processes stay waitable (cancel-safe). A non-zero exit or signal is returned as an Outcome, not an error. An empty slice resolves to an empty Vec.

Panics

Does not panic on any caller input: the final collection step carries an internal consistency assertion (every outcome slot is filled once all contenders have exited, an invariant the join loop maintains). It is documented only because that assertion is a hard expect.