Turns out the load-balancer we are using has a very short keepalive timeout, and our old value of 30 seconds was long enough for it to drop the connection. Setting this to 5 seconds on the client, and keeping 30 seconds for the signer until we have the signer notifications working again.
(cherry picked from commit 08b2accdcb59f4412e2842313f534d21a14d3e64)
Turns out the load-balancer we are using has a very short keepalive timeout, and our old value of 30 seconds was long enough for it to drop the connection. Setting this to 5 seconds on the client, and keeping 30 seconds for the signer until we have the signer notifications working again.
(cherry picked from commit 08b2accdcb59f4412e2842313f534d21a14d3e64)