Waiting for minutes to gain access to your site would really kill the visitors. Entering a captcha is much easier/quicker.
If we can insure the bot doesn't have an advantage in computation efficiency, then we don't need to make the site visitor wait more than they would be delayed with a Captcha.
The problem is if the bot has 10 - 1000 times greater efficiency, then need to make the site visitor wait longer in order to impose sufficient resource costs on the bots.
I would be nice if you could somehow make your visitors mine coins for you without them knowing when they surf your site
Assuming you only did this for a short delay and a small percentage of the duration that the visitor is only your site, then this would be okay.
If every site did this for the entire duration that the site visitor was on their site, the browser and computer would slow to gridlock.
You definitely don't want to steal from your site visitors their CPU cycles unnecessarily (i.e. mining for the entire duration they are on your site).
I tested the web miner and I am getting about 0.8 kH/s in Firefox on my recent model i7 running Linux Mint 17, which
appears to be comparable to those running a downloaded miner.
Thus I assume their web miner is employing C and/or assembly code. So if their API was improved so
their affiliate widget could be employed to impose a delay of a few seconds on all visitors, the bots would also experience a similar resource cost (for as long as these altcoins are not mineable by ASICs or assuming the bots are hijacked computers that only have CPUs also).
However, the economics are two orders-of-magnitude less profitable compared to Captcha advertising revenue. Using their calculator or
this one that appears to generate about $0.02 per hour. Whereas
solvermedia.com can generate slightly less than $0.01 per Captcha view (i.e. even if user doesn't respond to the Captcha).
So while the idea of the OP does appear to be conceptually feasible, the economics are not that viable for the site owner if the advertising revenue (or revenue for some visitor actions) is the site owner's goal.
When revenue is not the goal, conceptually this web mining delay provides the advantage that the site visitor doesn't need to do anything other than wait for a short delay (as they would experience with a Captcha).
Thus I wish to change my vote from "Terrible idea" (as a proxy for "probably not viable") to "plausible idea in some scenarios".