Keep US staging telemetry on for every build

README.md

@@ -51,7 +51,18 @@ progress JSON is uploaded to `policyengine/populace-us-staging` while the build
 runs (best-effort — a missing token or failed upload never fails the build), so
 every candidate shows up on the staging dashboard before it is published.
 Disable with `--no-staging`, or point elsewhere with `--staging-repo-id` /
-`POPULACE_STAGING_REPO_ID`. An *empty* `POPULACE_STAGING_REPO_ID` is ignored
+`POPULACE_STAGING_REPO_ID`. Uploads run on a background thread every 300 s
+(`--staging-upload-interval-seconds`), so a slow or unreachable Hub never
+stalls the build, and SIGTERM marks the run failed. Exact-count (`--exact-k`)
+builds stage too, under their own `runs/<run_id>/` prefix only: they never
+move `latest_staging.json` or `runs.json`. A build whose Hugging Face token
+cannot write the staging repository is never refused: it warns, keeps its
+telemetry under `runs/<run_id>/` locally, and builds as usual. Upload that
+folder later from any machine with a token:
+
+```bash
+uv run python tools/restage_us_staging_run.py --run-dir <release_root>/staging/runs/<run_id>
+``` An *empty* `POPULACE_STAGING_REPO_ID` is ignored
 rather than read as off, and staging with no destination at all is an argparse
 error — `--no-staging` is the only way a build produces no telemetry.
 

changelog.d/us-staging-on-every-build.changed.md

@@ -0,0 +1 @@
+US fiscal-refresh staging telemetry is now safe to leave on for every build. Exact-count builds stage under their own `runs/<run_id>/` prefix and never write `latest_staging.json` or `runs.json`, so `--exact-k` no longer requires `--no-staging` and the exact-k ladder launcher stops passing it. The release uploads from a background thread, so a slow or unreachable Hub never stalls the build; the terminal upload is awaited for at most 120 s. `--staging-upload-interval-seconds` defaults to 300 s (was 30 s), as for UK builds, because each cycle is up to six single-file Hub commits. SIGTERM now raises `BuildTerminatedError`, so a supervisor or budget stop marks the staging run failed instead of leaving it `running`. A build without a token that can write the staging repository is never refused: it warns at the start, keeps its telemetry local, records `delivery_check` in the run manifest, and builds as usual; `tools/restage_us_staging_run.py --run-dir <run folder>` uploads the run later from any machine with a token.