diff --git a/.github/workflows/canary.yml b/.github/workflows/canary.yml
index 4abc67b..f8f4a46 100644
--- a/.github/workflows/canary.yml
+++ b/.github/workflows/canary.yml
@@ -213,11 +213,18 @@ jobs:
 
           # A goal at or below zero, or a percentage computed from one, means
           # pct_funded's guards stopped guarding.
+          # Zero itself is a legitimate figure: a campaign that has raised
+          # 10,000 against a goal of 900,000,000 rounds to 0.0 at two
+          # decimals, and a campaign with nothing raised is 0% funded. What
+          # must never happen is a NEGATIVE percentage, or a percentage
+          # computed from a goal that is absent or not positive.
           bad_pct = [r.get("sku") for r in rows
-                     if r.get("pct_funded") is not None and r["pct_funded"] <= 0]
+                     if r.get("pct_funded") is not None
+                     and (r["pct_funded"] < 0
+                          or r.get("goal") is None or r["goal"] <= 0)]
           if bad_pct:
-              fail.append(f"{len(bad_pct)} row(s) report pct_funded <= 0: "
-                          f"{bad_pct[:3]}")
+              fail.append(f"{len(bad_pct)} row(s) report a negative pct_funded "
+                          f"or one without a positive goal: {bad_pct[:3]}")
 
           # A page number never threaded into the parser produces (page,
           # position) collisions across pages that look, from the row count
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 440c4eb..9ae476f 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -11,6 +11,19 @@ the release notes rather than left to be discovered from a bill or a diff.
 
 ## [Unreleased]
 
+### Fixed
+- The daily canary was red from 2026-10-07 to 2026-10-12 although the run
+  itself was healthy (72 rows, `complete`, 3 of 3 pages). Its sanity check
+  rejected `pct_funded <= 0`, and a real campaign that raised 10,000 against
+  a goal of 900,000,000 rounds to 0.0. The check now fails on a negative
+  percentage, or one computed from a goal that is not positive; zero is a
+  figure. The parser was right and is unchanged, with a check pinning the case.
+
+### Changed
+- `scraper_api_client.py` warns on stderr when the key is passed with `--key`
+  (visible to `ps`, kept in shell history). `TWOCAPTCHA_KEY` in `.env` is
+  unchanged and remains the documented way.
+
 > **Correction to v0.1.0's release notes.** They said the 2Captcha balance
 > was "unchanged to five decimal places". That was measured across the four
 > bare-browser runs and is true of them, but it is not true of the session as
diff --git a/README.md b/README.md
index f7866af..e3d2560 100644
--- a/README.md
+++ b/README.md
@@ -187,6 +187,11 @@ everything the site will serve. It is also a **sample**. The sidecar records
 `total_results`, `pages_available`, `capped_by_site` and `reachable_max` so
 a consumer can tell, and `diff_runs.py` says so before diffing.
 
+To read it: `capped_by_site: true` means the run holds a sample of what
+the query matched, however many pages it fetched. `pages_completed` against
+`pages_available` says how much of the *reachable* list the run covered, and
+`reachable_max` is the most rows the site will ever serve for it.
+
 A narrower query is not capped: `?term=solar` reported 801 matches with
 `hasCappedResults: false`.
 
diff --git a/scraper_api_client.py b/scraper_api_client.py
index 0114aad..9de54b2 100644
--- a/scraper_api_client.py
+++ b/scraper_api_client.py
@@ -479,6 +479,10 @@ def parse_args():
     p.add_argument("--dump-html", default=None,
                    help="Save the exact HTML the parser is given (always, even on success)")
     args = p.parse_args()
+    if args.key:
+        print("warning: --key puts the API key in this process's argv, where "
+              "anything that can run `ps` can read it, and in shell history. "
+              "Set TWOCAPTCHA_KEY in .env instead.", file=sys.stderr)
     # This client uses --key rather than --twocaptcha-key, so the env
     # mapping is spelled out instead of defaulted -- same pattern this
     # family's mediamarkt-scraper uses for its own scraper_api_client.py.
diff --git a/smoke_test.py b/smoke_test.py
index da07aa7..d03873e 100644
--- a/smoke_test.py
+++ b/smoke_test.py
@@ -748,6 +748,10 @@ def test_pct_funded_guards():
     assert pp.pct_funded(100.0, 0) is None
     assert pp.pct_funded(None, 100.0) is None
     assert pp.pct_funded(12576.05, 12000.0) == 104.8
+    # A real campaign (Aethersworn, 2026-10-12): 10,000 raised against a goal
+    # of 900,000,000 is 0.0 after rounding. It is a figure, not a missing
+    # value, and the daily canary once failed for six days by treating it as one.
+    assert pp.pct_funded(10000.0, 900000000.0) == 0.0
 
 
 @check("days_left is computed from the absolute end, not the localised card")