A "bad attempt" (i.e. a block whose hash is above the target) isn't "crossed out". It is entirely possible that you will later find a different block whose hash has that same value. Nothing prevents it. Every hash behaves like an independent trial.
It's easier to understand with smaller numbers. Say you have a 6-sided die and you want to roll it until you get a 6. If you roll a 3 on the first roll, that doesn't "cross out" the number 3; it is entirely possible that you will roll 3 again on future rolls. The die has no "memory". For that reason, there is no guarantee that you will get a 6 within 6 rolls; there's a chance it could be 7 rolls, or 12, or 100 rolls, before you get a 6.
Even if this were true, it wouldn't really make much difference: the number of hashes that you, or all the world combined, have ever computed, is a negligibly small fraction of the total number possible. 2^256 is an extremely large number.
http://bitcoin.sipa.be/ estimates that a total of 10^26 hashes have been performed on the Bitcoin network to date. That's 100000000000000000000000000. The total number of possible hashes is 115792089237316195423570985008687907853269984665640564039457584007913129639936. So about 0.0000000000000000000000000000000000000000000000001% of them have been seen so far. Even if they had been somehow "crossed out" it would not appreciably change the number that remained.
As to number 2, you are correct: since the goal is to find a hash that is less than the target, a smaller target means it is harder to find a successful hash. The number usually called "difficulty" is computed as the maximum possible target (which is 2^224) divided by the current target, so a smaller current target means a larger difficulty number.
Hi Nate, really appreciate your answer! I have a follow up question, What is the reason of not being able to "cross out" numbers? Is it because more than one input will produce the same hash so that even if you cross out a hash, there is no guarantee that the same hash will not appear again? – Cheng – 2017-02-19T01:18:36.013
1@Cheng: yes, that's right. – Nate Eldredge – 2017-02-19T01:29:07.820
According to this SO post, the rate of SHA256 collision is exetremely rare, but I guess as you stated, 2^256 is enormous, even if we assume there is no hash collision, it will still takes a lot computing power to find the right answer?
– Cheng – 2017-02-19T01:38:15.640Hi Nate, I asked a follow up question here if you are interested: http://bitcoin.stackexchange.com/questions/51474/how-many-correct-hashes-are-there-giving-a-difficulty
– Cheng – 2017-02-19T02:30:38.907