research: rerun cancels orphan runs first so 409 doesn't block restart

The previous rerun guard (standby OR (processing AND in_flight==0))
still 409'd when a lingering queued/running row existed — typically
an orphan from a server restart mid-pipeline, before the reaper or
stale-checkpoint requeuer had marked it failed.

Now: allow rerun on any 'processing' topic; before enqueuing the
fresh run, batch-cancel every in-flight run for the topic so the
new run isn't racing them. Terminal states (reviewing / publishing
/ published) still 409 as before.

crates/cm-api/src/routes/research.rs

@@ -679,16 +679,26 @@ pub async fn start_topic(
         .await?
         .ok_or(ApiError::NotFound)?;
     // Allow the fresh-start path (standby) AND the rerun path (topic
-    // parked in `processing` after all runs failed / no runs remain in
-    // flight). Blocks accidental double-fires on a live pipeline
-    // (in_flight > 0) and terminal states (reviewing / publishing /
-    // published).
-    let in_flight =
-        cm_db::repo::topology_runs::active_runs_for_research_topic(&state.pool, id).await?;
-    let can_start = topic.status == "standby" || (topic.status == "processing" && in_flight == 0);
+    // parked in `processing`). For rerun, first cancel any lingering
+    // queued/running runs (typically orphans left over from a server
+    // restart mid-pipeline) so the fresh enqueue isn't racing them.
+    // Terminal states (reviewing / publishing / published) still 409.
+    let can_start = topic.status == "standby" || topic.status == "processing";
     if !can_start {
         return Err(ApiError::Conflict);
     }
+    if topic.status == "processing" {
+        let orphan_ids =
+            cm_db::repo::topology_runs::active_run_ids_for_research_topic(&state.pool, id).await?;
+        for run_id in orphan_ids {
+            // Best-effort — cancel just marks the row; the actual worker
+            // (if still alive) stops at its next step boundary. If the
+            // worker is dead (server restart), the row transitions and
+            // stops being counted as in-flight immediately.
+            let _ =
+                cm_db::repo::topology_runs::cancel(&state.pool, run_id, user.workspace_id).await;
+        }
+    }
     // D1 fold — refuse to double-fire when a scheduled research loop
     // already owns this topic. Otherwise clicking the legacy "Start
     // research" button while a loop iteration is in flight would spawn