review: docs — fixture chunk wording, publish prefix results/, explicit hours unit

Author: marwen-abid

README.md

@@ -14,7 +14,7 @@ campaigns on an AWS NVMe devbox (`m6id.2xlarge`), driven by the config-driven ru
 [`runner/`](runner/) (see "Run a campaign" below). Each campaign is several configurations
 × 5 fresh-process runs; every run writes CSVs
 (`stage,n,n_items,total_ns,p50_ns,p90_ns,p99_ns,max_ns`) into its own directory, and the
-results are mirrored to GCS under `gs://rpc-full-history/benchmarks/`.
+results are mirrored to GCS under `gs://rpc-full-history/results/`.
 
 `converter/convert.py` (Python 3, standard library only) turns one results directory into
 `docs/runs/<run-id>.json` (schema v1) and updates the manifest `docs/runs/index.json`. The
@@ -86,7 +86,7 @@ go run ./cmd/campaign run my-campaign.toml
 #    bundle lacks survive it. Against a destination that really is empty,
 #    --force only skips the emptiness check, so it is safe on a first publish):
 go run ./cmd/campaign publish /mnt/nvme/bench/results/<run-id> \
-  gs://rpc-full-history/benchmarks --force
+  gs://rpc-full-history/results --force
 ```
 
 The published bundle is exactly what the next section ingests into a committed run

runner/README.md

@@ -143,8 +143,9 @@ that root is materialized:
   `$BENCH_ROOT/golden/<name>/` and reused by later campaigns.
 - **`bsb-s3`** — `location` is an S3 bucket path; an untimed cold backfill materializes
   `$BENCH_ROOT/golden/<name>/`, one chunk at a time.
-- **`fixture`** — no location and no network: `ledgers` chunks are generated, then an
-  untimed cold ingest freezes them into `$BENCH_ROOT/golden/<name>/`.
+- **`fixture`** — no location and no network: each chunk in `chunks` is generated with
+  `ledgers` ledgers, then an untimed cold ingest freezes them into
+  `$BENCH_ROOT/golden/<name>/`.
 
 Anything the runner materializes is built under `<root>.partial` and renamed onto
 `<root>` only once whole, so an interrupted preparation is never mistaken for a finished
@@ -244,7 +245,7 @@ root, survives an instance stop) and, when `publish_uri` is set, uploaded to
 ## Resuming a crashed campaign
 
 A campaign is hours of work — the phase-1 reference run took ~17 hours, with single
-hot-ingest legs near 5.5 — so a crash or an OOM kill at the last rep should not cost the
+hot-ingest legs near 5.5 hours — so a crash or an OOM kill at the last rep should not cost the
 whole thing. `--resume` continues into the existing results directory instead of starting
 a new one:
 

runner/example-campaign.toml

@@ -65,7 +65,7 @@ query = false
 # Publish the finished bundle to object storage (gs:// or s3://) at the end of
 # the campaign. Default "" = keep results local; publish later by hand with
 # `campaign publish <bundle> <dest-root>`.
-#publish_uri = "gs://rpc-full-history/benchmarks"
+#publish_uri = "gs://rpc-full-history/results"
 
 # Datasets to benchmark: one [[dataset]] table each, with a unique name and the
 # chunk IDs this campaign runs. Every kind converges on a local cold pack root