Thanks for the links. I was studying them yesterday, and they have certainly helped. I now am wondering if I am overthinking this.
I am a bit puzzled about the concept of lag, for instance. I understand how it seems to work from the second text, and in a static, statistical sense I can understand it, but in a streaming system I don't see why you would ever need to calculate it for anything other than a lag of zero? (since the stream itself will slide across the reference signal as it is received.) Am I missing something?
It would seem that with each sample a new correlation value is generated, which could then be used to look for trends, and from that an edge or peak detection is used detect "maximal correlation".