How does the incremental crawl know what are the recent changed documents? If by checking the SP modified date column, then what about external and non SharePoint sources and html web sites?
The incremental process is dependent on the protocol handler being used. When detecting changes against Sharepoint sites, we use the "Sharepoint 3.0 (sts3) protocol handler".
We will first attempt to get changes from the last crawl. We do this through MSDMN.exe process and hit a webservice called sitedatawebservice.
Read this article for detailes(this blog explain about Sitec changes).