I don't think this can be done. If you use the weblogic proxy plugin and address a comma separated list of managed server as a cluster using the WeblogicCluster element (https://docs.oracle.com/middleware/1212/webtier/PLGWL/apache.htm#PLGWL406 ), then OHS can sense or be signaled that a managed server is "in the air" or not.
However, if you get a HTTP 404 or a 50x exception then basically the server is up, as far as OHS is concerned.
OHS can't know that the 404 is because you address a resource that really does not or isn't loaded by the SOA Infrastructure (it actually does not know that the weblogic is a SOASuite Server). Also if you get an internal server error: it actually means that the resource is there, but failed because of an internal exception. Is that because you provided a request that did not conform the prerequisites of the service or if soa-infra isn't done loading? How would OHS know?
As far as initial startup counts: you probably should only start the OHS after the SOASuite is fully up. For a restart/failover of a single node within a cluster I don't have a proper idea for now.