Reset reconnection attempt counter when server makes progress

When the server sends new events before closing the stream (indicated by
a changed last_event_id), the reconnection is intentional and the attempt
counter should reset to 0. Only increment the counter when no new events
were received, which indicates an actual failure to make progress.

This prevents legitimate server-initiated reconnection patterns from
being penalized by the retry limit while still protecting against
infinite loops when the server returns empty responses.

Github-Issue: #2397

Author: Christian-Sidak

src/mcp/client/streamable_http.py

@@ -423,7 +423,10 @@ async def _handle_reconnection(
 
                 # Stream ended again without response - reconnect again
                 logger.info("SSE stream disconnected, reconnecting...")
-                await self._handle_reconnection(ctx, reconnect_last_event_id, reconnect_retry_ms, attempt + 1)
+                # Reset attempt counter if server made progress (sent new events),
+                # otherwise count as a failed attempt
+                next_attempt = 0 if reconnect_last_event_id != last_event_id else attempt + 1
+                await self._handle_reconnection(ctx, reconnect_last_event_id, reconnect_retry_ms, next_attempt)
         except Exception as e:  # pragma: no cover
             logger.debug(f"Reconnection failed: {e}")
             # Try to reconnect again if we still have an event ID