run_one: exactly 1 attempt (adapter handles all retries internally)

Co-Authored-By: Claude <noreply@anthropic.com>

evalharness/model/runner.py

@@ -419,18 +419,15 @@ async def generate_predictions(
         # backoff so one ConnectError burst cannot kill the whole batch --
         # the adapter already retries 5xx/429 and the pool fails over per
         # instance; this is the last line of defense around asyncio.gather
-        for attempt in range(3):
+        # NO outer retry: the adapter retries internally; a second loop
+        # here multiplied worst-case time (42+ attempts before this fix).
+        # One pass, one result or one error.
         try:
             pred = await one(s)
-                break
         except Exception:
-                if attempt == 2:
             if progress_reporter is not None:
                 progress_reporter.advance(success=False)
             raise
-                # minute-scale backoff: cluster routes flap in multi-minute
-                # bursts; short retries exhaust inside one dead window
-                await asyncio.sleep((10, 20, 40)[attempt])
         if ckpt_store is not None:
             ckpt_store.append(keys[i], pred)
         return i, pred