diff --git a/tests/lifecycle/bats/all-apps-lifecycle.bats b/tests/lifecycle/bats/all-apps-lifecycle.bats
index fce49404..d35c4c6b 100644
--- a/tests/lifecycle/bats/all-apps-lifecycle.bats
+++ b/tests/lifecycle/bats/all-apps-lifecycle.bats
@@ -129,11 +129,21 @@ catalog_install_payload() {
   local fails="" id
   for id in $(target_apps); do
     [[ "$(app_state "$id")" == "running" ]] || continue   # only cycle running apps
-    rpc_result package.stop "{\"id\":\"$id\"}" >/dev/null 2>&1
+    # Each rpc_result must be allowed to fail without aborting the loop.
+    # rpc_result returns non-zero whenever the response carries .error, and
+    # under bats' errexit a bare call ends the test right there — so a single
+    # transient RPC hiccup killed the run BEFORE the $fails summary below could
+    # name the app. That is exactly what happened on 2026-08-08: the whole test
+    # died at package.stop with no indication of which of the ten targets it was
+    # (it was mempool, and the same call succeeded by hand moments later).
+    rpc_result package.stop "{\"id\":\"$id\"}" >/dev/null 2>&1 \
+      || { fails+="$id:stop-rpc "; continue; }
     wait_state "$id" stopped 120 || { fails+="$id:stop "; }
-    rpc_result package.start "{\"id\":\"$id\"}" >/dev/null 2>&1
+    rpc_result package.start "{\"id\":\"$id\"}" >/dev/null 2>&1 \
+      || { fails+="$id:start-rpc "; continue; }
     wait_state "$id" running 240 || { fails+="$id:start "; continue; }
-    rpc_result package.restart "{\"id\":\"$id\"}" >/dev/null 2>&1
+    rpc_result package.restart "{\"id\":\"$id\"}" >/dev/null 2>&1 \
+      || { fails+="$id:restart-rpc "; continue; }
     wait_state "$id" running 240 || { fails+="$id:restart "; }
   done
   [[ -z "$fails" ]] || { echo "# lifecycle failures: $fails" >&3; false; }