1 - there are two versions of truth in the network, the next miner to mine the next block will determine which previous block will be appended to the blockchain based on the one it chooses.
2 & 3 - It is a block that is left from a situation where there were two blocks solved at the same time and the next miner has chosen the other block, one block (the longest) gets appended to the blockchain and the other is a stale block
4 - to minimise the risk of your block being dropped because of the situation mentioned above