I can't speak for DNSDumpster, but a common technique I use to do subdomain enumeration is just brute forcing with a wordlist. By enumerating with a large enough wordlist, you can discover matching subdomains for a target domain.
Yeah this is a random string of characters so it can't be that.. Also my logs show it getting queried by automated tools within a few weeks of being live, so I suspect it's due to someone's else's suggestion that a DNS resolver somewhere is caching and sharing request data for "research" or other reasons.
