High Latency OpenSearch Behind ALB

I am trying to set up OpenSearch in a private VPC subnet behind a Load Balancer in a public subnet. The load balancer endpoint is in turn placed in a Cloudfront distribution. Right now I am testing this with HTTP -- will try HTTPS once we are able to set up our DNS. After configuring the security groups to allow OpenSearch and the ALB to communicate, and after adding the listener/target group, I am able to connect to OpenSearch through the load balancer endpoint. However, if I try to access via the Cloudfront endpoint, I get a 504 Error: The Request Could Not Be Satisfied. I try pinging the ALB endpoint via curl and notice that it is taking 75 seconds to respond with 200-OK. So it seems that Cloudfront is not responding due to late responses from the Load Balancer. It always takes exactly 75 seconds -- except sometimes when I fire up the cluster, the first response comes back in a fraction of a second as it should, then on all subsequent attempts it takes 75 seconds. I am in Maryland and the cluster is set up in the Oregon region. I tried this with three progressively larger instances of OpenSearch and the compute power made no difference. I've been trying to figure this out for weeks -- any suggestions on what I am doing wrong? Thanks!