diff --git a/scripts/fleet_monitor.sh b/scripts/fleet_monitor.sh
index 01b2c4aa2..bb30d9dc3 100755
--- a/scripts/fleet_monitor.sh
+++ b/scripts/fleet_monitor.sh
@@ -267,6 +267,12 @@ fi
 # repos with >=1 marker in the window get a row; the section is omitted entirely
 # when the fleet had no dev-lead activity. Best-effort: a comment-read failure on
 # one repo is a per-repo warning, not a fatal error for the monitor.
+#
+# The timeout rate is reported as prevalence (timeout / total dev-lead runs, #1030),
+# so each row also carries the repo's total `dev-lead.yml` runs in the window —
+# read from the per-workflow metrics already collected above (matched by repo +
+# workflow basename), not a new API scan. A missing row or non-numeric total (the
+# ERROR sentinel `?`) sanitizes to 0, which the report renders as `n/a`.
 # ---------------------------------------------------------------------------
 dev_lead_reason_file=$(mktemp)
 for repo in "${repos[@]}"; do
@@ -280,11 +286,33 @@ for repo in "${repos[@]}"; do
   IFS=$'\t' read -r dl_timeout dl_engine_error dl_total \
     < <(summarize_dev_lead_timeouts "$comments_json")
   if [ "${dl_total:-0}" -gt 0 ]; then
-    printf '%s\t%s\t%s\t%s\n' "$repo" "$dl_timeout" "$dl_engine_error" "$dl_total" \
+    # Total dev-lead runs = the `dev-lead.yml` row's `total` (field 4) for this
+    # repo in metrics_file, matched by workflow-file basename. Empty when the
+    # repo has no dev-lead.yml row; '?' when that row is an ERROR sentinel.
+    dl_runs=$(awk -F'\t' -v r="$repo" '
+      $2 == r {
+        n = split($3, parts, "/")
+        if (parts[n] == "dev-lead.yml") { print $4; exit }
+      }' "$metrics_file")
+    case "${dl_runs:-}" in
+      ''|*[!0-9]*) dl_runs=0 ;;
+    esac
+    printf '%s\t%s\t%s\t%s\t%s\n' \
+      "$repo" "$dl_timeout" "$dl_engine_error" "$dl_total" "$dl_runs" \
       >> "$dev_lead_reason_file"
   fi
 done
 
+# Fleet-wide total dev-lead runs in the window = the sum of every repo's
+# `dev-lead.yml` row `total` (field 4) in the per-workflow metrics, across ALL
+# repos — including all-success repos with no markers. This is the prevalence
+# denominator for the timeout rate (#1030); summing only marker repos would
+# inflate the rate. Non-numeric totals (the ERROR sentinel `?`) are skipped.
+fleet_dl_runs=$(awk -F'\t' '
+  { n = split($3, p, "/"); if (p[n] == "dev-lead.yml" && $4 ~ /^[0-9]+$/) s += $4 }
+  END { print s + 0 }
+' "$metrics_file")
+
 # ---------------------------------------------------------------------------
 # 3. Generate reports
 # ---------------------------------------------------------------------------
@@ -311,7 +339,7 @@ stub_drift_section() {
 dev_lead_timeout_section() {
   [ -s "$dev_lead_reason_file" ] || return 0
   printf '\n'
-  generate_dev_lead_timeout_report "$dev_lead_reason_file"
+  generate_dev_lead_timeout_report "$dev_lead_reason_file" "${fleet_dl_runs:-0}"
 }
 
 # Step Summary — Tier 1 visualizations only (Mermaid not rendered there)
diff --git a/scripts/fleet_report.sh b/scripts/fleet_report.sh
index 13ec7068e..9c18eb2e0 100755
--- a/scripts/fleet_report.sh
+++ b/scripts/fleet_report.sh
@@ -148,40 +148,63 @@ summarize_dev_lead_timeouts() {
 
 # generate_dev_lead_timeout_report <reason_tsv_file>
 # Renders the Dev-Lead timeout ("wall") observability section (#1019, story C of
-# #901). Input is a TSV file with one row per repo that had >=1 dev-lead marker
-# in the window: `repo <TAB> timeout <TAB> engine_error <TAB> total`. Prints a
-# per-repo table, fleet totals, and the timeout rate (timeouts / all dev-lead
-# failure markers) — the "wall rate" Story A/B aim to drive down. An empty or
-# missing file prints nothing, so the section is omitted when there was no
-# dev-lead activity in the window.
+# #901; prevalence rate #1030). Input is a TSV file with one row per repo that
+# had >=1 dev-lead marker in the window:
+#   `repo <TAB> timeout <TAB> engine_error <TAB> markers <TAB> total_runs`
+# where `markers` is the count of all dev-lead failure/retry/escalation markers
+# (the composition breakdown) and `total_runs` is the repo's total dev-lead.yml
+# runs in the window (the prevalence denominator, from the fleet monitor's
+# per-workflow metrics). Prints a per-repo table (raw timeout + engine-error
+# counts retained), fleet totals, and the timeout **prevalence** rate
+# (`reason=timeout` / total dev-lead runs — over ALL runs, not just failures) —
+# the "wall rate" Story A/B aim to drive down. When total_runs is missing,
+# non-numeric (e.g. an ERROR sentinel), or zero for a repo it is shown as `n/a`
+# and contributes 0 to the denominator (no divide-by-zero). An empty or missing
+# file prints nothing, so the section is omitted when there was no dev-lead
+# activity in the window.
 generate_dev_lead_timeout_report() {
-  local f="${1:-}"
+  local f="${1:-}" fleet_runs="${2:-}"
   [ -n "$f" ] && [ -s "$f" ] || return 0
+  # Prevalence denominator = ALL dev-lead runs fleet-wide (passed in by the
+  # caller from the per-workflow metrics), NOT just runs from repos that happened
+  # to have a marker. Summing only marker repos drops every all-success repo from
+  # the denominator and inflates the rate — the exact flaw #1030 set out to fix.
+  case "${fleet_runs:-}" in ''|*[!0-9]*) fleet_runs=0 ;; esac
 
-  local repo t e tot
+  local repo t e tot runs runs_disp
   local sum_t=0 sum_e=0 sum_tot=0
 
   printf '## Dev-Lead Timeouts (walls)\n\n'
   printf 'Stage timeouts (`reason=timeout`) that escalate to a human, broken out from generic `engine-error`, over the window.\n\n'
-  printf '| Repo | Timeouts | Engine-errors | Total dev-lead failures |\n'
-  printf '|---|---|---|---|\n'
-  while IFS=$'\t' read -r repo t e tot; do
+  printf '| Repo | Timeouts | Engine-errors | Failure markers | Total runs |\n'
+  printf '|---|---|---|---|---|\n'
+  while IFS=$'\t' read -r repo t e tot runs; do
     [ -n "$repo" ] || continue
-    printf '| `%s` | %s | %s | %s |\n' "$repo" "$t" "$e" "$tot"
+    # Sanitize total_runs: empty / non-numeric (ERROR sentinel '?') → 0 → n/a.
+    case "${runs:-}" in
+      ''|*[!0-9]*) runs=0 ;;
+    esac
+    runs_disp="n/a"
+    [ "$runs" -gt 0 ] && runs_disp="$runs"
+    printf '| `%s` | %s | %s | %s | %s |\n' "$repo" "$t" "$e" "$tot" "$runs_disp"
     sum_t=$(( sum_t + ${t:-0} ))
     sum_e=$(( sum_e + ${e:-0} ))
     sum_tot=$(( sum_tot + ${tot:-0} ))
   done < "$f"
-  printf '| **Fleet total** | **%s** | **%s** | **%s** |\n\n' "$sum_t" "$sum_e" "$sum_tot"
+  local runs_total_disp="n/a"
+  [ "$fleet_runs" -gt 0 ] && runs_total_disp="$fleet_runs"
+  printf '| **Fleet total** | **%s** | **%s** | **%s** | **%s** |\n\n' \
+    "$sum_t" "$sum_e" "$sum_tot" "$runs_total_disp"
+  printf '_Per-repo rows list only repos with >=1 dev-lead marker in the window; the **Total runs** fleet total (and the rate denominator) include **all** dev-lead runs fleet-wide, incl. all-success repos with no markers._\n\n'
 
   local rate
-  rate=$(awk -v t="$sum_t" -v n="$sum_tot" 'BEGIN {
+  rate=$(awk -v t="$sum_t" -v n="$fleet_runs" 'BEGIN {
     if (n <= 0) { print "n/a"; exit }
     pct = t * 100 / n
     printf (pct == int(pct)) ? "%d%%" : "%.1f%%", pct
   }')
-  printf '**Timeout rate** (`reason=timeout` / all dev-lead failure markers): %s (%s / %s)\n' \
-    "$rate" "$sum_t" "$sum_tot"
+  printf '**Timeout rate** (`reason=timeout` / total dev-lead runs fleet-wide — prevalence over all runs, not just failures): %s (%s / %s)\n' \
+    "$rate" "$sum_t" "$runs_total_disp"
 }
 
 # detect_systemic_failures <metrics_file>
diff --git a/tests/fleet_report.bats b/tests/fleet_report.bats
index c9b8c1f62..2f05106b9 100755
--- a/tests/fleet_report.bats
+++ b/tests/fleet_report.bats
@@ -405,44 +405,100 @@ _mk_metrics() {
 }
 
 # ---------------------------------------------------------------------------
-# generate_dev_lead_timeout_report — renders the per-repo section (#1019)
-# Input: TSV file rows `repo <TAB> timeout <TAB> engine_error <TAB> total`.
+# generate_dev_lead_timeout_report — renders the per-repo section (#1019/#1030)
+# Input: TSV file rows
+#   `repo <TAB> timeout <TAB> engine_error <TAB> markers <TAB> total_runs`.
+# The timeout rate is prevalence: timeout / total_runs (over ALL dev-lead runs,
+# not just failure markers).
 # ---------------------------------------------------------------------------
 
-@test "dev-lead timeout report: renders per-repo rows, fleet total, and timeout rate" {
+@test "dev-lead timeout report: renders per-repo rows, fleet total, and prevalence rate" {
   local f
   f="$(mktemp "$BATS_TEST_TMPDIR/fr.XXXXXX")"
+  # timeouts 2+1=3; total runs 60+40=100 → prevalence 3/100 = 3%
   printf '%s\n' \
-    $'petry-projects/a\t2\t1\t4' \
-    $'petry-projects/b\t1\t1\t3' \
+    $'petry-projects/a\t2\t1\t4\t60' \
+    $'petry-projects/b\t1\t1\t3\t40' \
     > "$f"
-  run generate_dev_lead_timeout_report "$f"
+  run generate_dev_lead_timeout_report "$f" 100
   [ "$status" -eq 0 ]
   [[ "$output" =~ "petry-projects/a" ]]
   [[ "$output" =~ "petry-projects/b" ]]
-  # Fleet totals: timeout 3, engine-error 2, total 7
+  # Fleet totals: timeout 3, engine-error 2, markers 7 (composition breakdown retained)
   [[ "$output" =~ "Fleet total" ]]
   [[ "$output" =~ "reason=timeout" ]]
-  # Rate = 3 / 7 = 42.9%
-  [[ "$output" =~ "42.9%" ]]
+  # Prevalence rate = K/N = 3 / 100 = 3% (over total runs, not failures)
+  [[ "$output" =~ "3%" ]]
+  [[ "$output" =~ "3 / 100" ]]
+  # Not the old composition rate (3 / 7 = 42.9%)
+  [[ ! "$output" =~ "42.9%" ]]
+  rm -f "$f"
+}
+
+@test "dev-lead timeout report: denominator is ALL fleet runs, not just marker repos (#1030)" {
+  local f
+  f="$(mktemp "$BATS_TEST_TMPDIR/fr.XXXXXX")"
+  # Only repo a has markers (2 timeouts over its 10 runs). But the fleet ran 100
+  # dev-lead runs total — repo b had 90 all-success runs (no markers, absent from
+  # the file). Prevalence MUST be 2/100 = 2%, not 2/10 = 20% (the marker-repo-only
+  # denominator bug). The passed-in fleet total is the denominator.
+  printf '%s\n' $'petry-projects/a\t2\t0\t2\t10' > "$f"
+  run generate_dev_lead_timeout_report "$f" 100
+  [ "$status" -eq 0 ]
+  [[ "$output" =~ "2 / 100" ]]
+  [[ ! "$output" =~ "20%" ]]
+  rm -f "$f"
+}
+
+@test "dev-lead timeout report: prevalence rate K/N renders a decimal when needed" {
+  local f
+  f="$(mktemp "$BATS_TEST_TMPDIR/fr.XXXXXX")"
+  # K=3 timeouts over N=8 total runs → 37.5%
+  printf '%s\n' $'petry-projects/a	3	0	3	8' > "$f"
+  run generate_dev_lead_timeout_report "$f" 8
+  [ "$status" -eq 0 ]
+  [[ "$output" =~ "37.5%" ]]
+  [[ "$output" =~ "3 / 8" ]]
   rm -f "$f"
 }
 
 @test "dev-lead timeout report: whole-number rate renders without a decimal" {
   local f
   f="$(mktemp "$BATS_TEST_TMPDIR/fr.XXXXXX")"
-  printf '%s\n' $'petry-projects/a\t1\t1\t2' > "$f"
-  run generate_dev_lead_timeout_report "$f"
+  # 1 timeout over 2 total runs → 50%
+  printf '%s\n' $'petry-projects/a	1	1	2	2' > "$f"
+  run generate_dev_lead_timeout_report "$f" 2
   [[ "$output" =~ "50%" ]]
   [[ ! "$output" =~ "50.0%" ]]
   rm -f "$f"
 }
 
+@test "dev-lead timeout report: zero total runs yields n/a (no divide-by-zero)" {
+  local f
+  f="$(mktemp "$BATS_TEST_TMPDIR/fr.XXXXXX")"
+  printf '%s\n' $'petry-projects/a	1	0	1	0' > "$f"
+  run generate_dev_lead_timeout_report "$f" 0
+  [ "$status" -eq 0 ]
+  [[ "$output" =~ "n/a" ]]
+  rm -f "$f"
+}
+
+@test "dev-lead timeout report: unavailable (non-numeric) total runs is treated as n/a" {
+  local f
+  f="$(mktemp "$BATS_TEST_TMPDIR/fr.XXXXXX")"
+  # 5th column is the ERROR sentinel '?' — must not crash or divide-by-zero
+  printf '%s\n' $'petry-projects/a	1	0	1	?' > "$f"
+  run generate_dev_lead_timeout_report "$f" '?'
+  [ "$status" -eq 0 ]
+  [[ "$output" =~ "n/a" ]]
+  rm -f "$f"
+}
+
 @test "dev-lead timeout report: empty file produces no output" {
   local f
   f="$(mktemp "$BATS_TEST_TMPDIR/fr.XXXXXX")"
   : > "$f"
-  run generate_dev_lead_timeout_report "$f"
+  run generate_dev_lead_timeout_report "$f" 0
   [ "$status" -eq 0 ]
   [ -z "$output" ]
   rm -f "$f"