**Time each source separately.** Add timing logs per source in the compilation run and run it once. Expected: one or two sources account for most of the total wait, instead of a uniform slowdown.

**Add a per source timeout.** Wrap each source fetch in a timeout well below the job limit, and mark that source as skipped on timeout. Success check: a hanging source no longer blocks the rest of the briefing.

**Fetch in parallel.** Fetch independent sources concurrently with a sensible connection limit for the host machine. Expected: total fetch time is close to the slowest single source, not the sum of all sources.

**Add a fallback cache.** Keep the last good result per source and use it, clearly dated, when todays fetch times out. Success check: the briefing still contains that section instead of aborting.

**Publish partial, report missing.** Build the briefing from the sources that succeeded and add a short note listing any source that was skipped. Expected: the briefing is delivered on time even on a bad source day.

**Verify on a normal run.** Run the scheduled job again end to end. Success check: it finishes within its window and the delivery step runs.