[g_sawchuk]

I would confidently say that it is definitely beneficial to try and counter a 2-boulder autonomous with a "steal". The plausibility is definitely present. You need to intake a ball in the time it takes them to:

A. Deposit their pre-load.
B. Intake the ball.

Without a doubt, even with a slightly less efficient robot, you could beat them to the boulder.

Having now confirmed that the strategy is plausible, we want to look at the value of stealing the ball. Assuming that you plan on shooting your boulder too, and you take a boulder, we're looking at the following scores (under the assumption that no other members of the alliance do any scoring tasks).
We will call our robot, the "stealer", the BLUE alliance, and the two-boulder-autonomous robot the RED alliance.

BLUE: 30
RED: 30

Stealing their targeted boulder effectively accomplishes the following:

A. Evens the scores.
B. You get a boulder out of the neutral zone, forcing it onto their side, making the boulder more accessible, as opposed to being way at the other side of the field after they have scored them.

Now, let's assume that the BLUE alliance robot is now only capable of intaking, and will only steal the boulder, and not breach or score the boulder. We would now see the following scores appear.

BLUE: 0
RED: 20

The margin is now greater (20 points), but they have only been able to minimize the margin. The BLUE alliance has benefited themselves by:

A. Reducing the score margin by 10 points.
B. Reduced the number of strength that their castle has lost in autonomous.
C. Gaining possession of one of the neutral zone boulders, limiting how many "easy cycles" the other alliance can complete by grabbing the boulders from the neutral passage, typically shortening their cycles.

From this, we can evidently see that even the simplest ball steal is quite valuable, and, even if it isn't a game-changer, it can still minimize the score margin.