Zsh prompt latency and background job notification limits

Answer

Yes. Zsh has a built-in, non-blocking mechanism for job completion alerts, and precmd job polling can be throttled to avoid prompt lag.

Confirmed facts

• setopt notify is a documented Zsh option. With it enabled the shell prints a job completion message as soon as the job terminates, driven by Zsh's internal SIGCHLD handling. No user code runs in the main thread to poll.
• precmd runs synchronously before each primary prompt is drawn. Any work done there, e.g. jobs, jobs -p, or looping over $jobstates, adds directly to ZLE prompt latency.

Likely explanation for your lag

Polling background PIDs on every precmd is O(number of jobs) per prompt. With many jobs the jobs builtin and any custom status formatting become perceptible. This is a common cause of prompt slowdown, not a Zsh bug.

Steps for this case

1. Use SIGCHLD-driven alerts instead of polling for immediate notification.

   setopt notify

   With notify a message like [1] Done sleep 2 is printed immediately on exit, without waiting for the next prompt. This removes the need for a precmd check to provide timely feedback.

2. Keep a minimal precmd guard if you still want a periodic summary.

   LAST_JOB_CHECK=0
   JOB_CACHE=""
   
   precmd_job_guard() {
     local now=$EPOCHREALTIME
     # throttle to ~1s
     (( ${now%.*} - ${LAST_JOB_CHECK%.*} < 1 )) && return
     LAST_JOB_CHECK=$now
   
     local current=$(jobs -p)
     [[ $current == $JOB_CACHE ]] && return
     JOB_CACHE=$current
   
     # your lightweight summary here, avoid heavy formatting
   }
   add-zsh-hook precmd precmd_job_guard

   The guard skips work when the prompt arrives quickly and skips work when the PID set has not changed.

3. Reduce work inside precmd.
   • Avoid calling jobs with formatting options on every prompt.
   • Prefer reading $jobstates which is maintained by Zsh, rather than spawning external commands.
   • Move expensive rendering to preexec or run it asynchronously with (... &!) if you must keep it.

Combined approach: setopt notify for instant, SIGCHLD-driven alerts, and a throttled, cache-guarded precmd for any periodic UI. This keeps the prompt responsive while still surfacing job completion.

Verification you can do locally: start a new Zsh with setopt notify, run sleep 2 &, and confirm the Done message appears at ~2s without pressing Enter. Then add the timestamp guard and run several background jobs to confirm prompt return is immediate.

Note on behavior: notify interleaves messages with terminal output, which some users find noisy. Throttling interval trade-off: intervals longer than ~1s can miss very short jobs in the precmd summary, though notify will still print them.

One missing diagnostic that would change the recommendation: which prompt framework and exact precmd implementation you are using. If it already uses jobs with heavy formatting or external commands, the specific commands to replace will differ.