"retry_on_error on" not effective for failed CONNECT request

[maeder+squid]

Hi

I have a setup with a squid, say S, used by the clients. S is configured to have two parent squids, say P1 and P2, of which P1 has the higher weight.

retry_on_error is configured to on, causing GET and POST requests to be retried via P2 after having failed via P1.

Unfortunately, CONNECT requests are currently *not* retried via P2 after having failed (i.e. resulted in a connect timeout or a non-2xx status code)
via P1.

Is there a way to configure S so that CONNECT requests are retried the same way GET and POST requests already are?

Thanks
--
Thomas