After first review, the patch is not correct, because the size of the data to be added in the connection queue results in going over the limit of `tcpconn_wq_max` (corresponding to config parameter `tcp_conn_wq_max`).
If you have a different opinion, can you provide more details about how you concluded this would be a fix?