In high volume systems, low probability events happen very frequently.
What makes this a massive effort? I assume that GoDaddy has existing software to register a domain name given the name, and some fixed configuration like who should be listed as owner, contact, etc. So we need to call that logic, passing as input each domain that has been searched for.
Upon every search for a domain name, enqueue the search term to a queue. (This can happen e.g. after the page load completes.) The search results already indicate whether the name is registered, so only write unregistered domains to the queue.
Set up a fleet of workers to pull items from that queue and call the functionality that registers a domain.
It seems feasible to have a basic prototype of this functionality operational within a day or two of effort (e.g., a dumb implementation that simply registers every search term that's a valid domain). I'm sure the production version will need to be smarter, like not registering every single search, or estimating the value of a prospect and only registering high-value prospects - but wouldn't a basic flagrant implementation be fairly simple?
Well for one, if you have ever worked on a well established high traffic site, even to add a "simple feature" without breaking something is not that simple.
Going with your logic: User Searching for a domain; Godaddy adds it to the queue; Pulls out the domains one by one from the queue and registers it for themselves. Although it might seem simple there are few flaws in it.
a) What happens if the user want to register the available domain, wait but we already added it to our queue for automatic registration. So we need to check if the user goes through with the registration or not.
b) We are going to make millions $ holding domains hostages. But since every registration will cost goddaddy atleast $5.50 to ICANN and that there are millions of searches of domains in a month, well it certainly is one way to tank your company.
c) Alright lets then try to filter out only the popular domain name searches. But how do we determine what is popular and what is not popular. Common words? Trends? Hyphenation? Numbers? TLD? User Profiling?
d) Lets hold the domain hostage and make the user wait until they agree to backorder it from us. That will always work out for user retention.
Given that I used to own a lot of domains from Godaddy, I can attest to the fact that they have much simpler ways to hold your domains hostage if they wanted to. For example: When a domain is about to expire, the sheer volume of emails you get from Godaddy reminding you to renew is outright annoying. If they had to score better money, cut it down to just one and sent them about 4 weeks before the expiration. When most people forget to do it, go ahead and charge them $80 to get it back from redemption!!
This isn't true, at least as of Dec. 2008: http://domainnamewire.com/2008/12/03/standard-tactics-llc-ho... .
But even without that I'm hugely skeptical of your "massive effort" claim.
I'm trying to believe this is just rotten luck, but every single account of this happening seems to link back to GoDaddy.