I would strongly recommend looking in more detail into available resources for SD and Kozak sequences, wikipedia basically answers these questions and has plenty of further reading if you desire to explore these questions.
At the same time, remember that these are statistical processes involving thousands of molecules, rather than deterministic processes happening at a single molecule. Thus, what changes with different sequences is a rate at which some step occurs (e.g. rate of ribosome assembly, rate of translation initiation, rate of start recognition); none of these steps will occur with 100% probability on every mRNA molecule.
What this means is that you can write down equations for kinetic models of translation, allowing us to apply principles of mass action and rates of reaction from elementary chemistry. Having a higher rate of ribosome affinity for a sequence is therefore similar in effect to just having more mRNA, in the sense that in both cases you will do more translation and make more protein.
Thus, it might be easier to use the terminology of "strong" and "weak" SD/Kozak sequences. A strong sequence has a high rate of reaction and a weak sequence has a low rate of reaction. Here, the "reaction" in question is ribosome binding, or translation initiation, or whatever process we're currently studying.
Now, as to the specific questions:
- Sequences alter the rate at which downstream processes occur. So, e.g. a weak Kozak sequence will often simply not initiate translation, and the ribosome will fall off without translating anything. This will almost never happen with a strong Kozak sequence, so a strong Kozak is more "efficient" in the sense of more reliably leading to translation. A strong Shine-Dalgarno sequence increases the rate at which ribosome binding occurs.
Both ribosome binding and translation initiation are necessary for protein production, so if you increase the rate at which these processes occur, you increase translational efficiency in each case.
- There is no "foreseeing". Think about this mechanistically. The ribosome scans 3' along the mRNA, starting at the point of attachment. Also, each ribosome attachment to an mRNA can only start one translation event. Once a ribosome stops scanning and starts translating, it does not recognize subsequent Kozaks.
Thus, if there is a strong Kozak sequence, then translation will almost always start at the strong sequence, and will never have the opportunity to start at a sequence further 3'. It is mostly only when a 5' Kozak is weak that further Kozaks get recognized at all ("leaky scanning"):
The first start codon closest to the 5′ end of the strand is not always recognized if it is not contained in a Kozak-like sequence. Lmx1b is an example of a gene with a weak Kozak consensus sequence.[22] For initiation of translation from such a site, other features are required in the mRNA sequence in order for the ribosome to recognize the initiation codon. Exceptions to the first AUG rule may occur if it is not contained in a Kozak-like sequence. This is called leaky scanning and could be a potential way to control translation through initiation.[23] For initiation of translation from such a site, other features are required in the mRNA sequence in order for the ribosome to recognize the initiation codon. (Wikipedia)
So it's not that Kozaks are required, necessarily. You just need some set of circumstances which increase the probability that the ribosome starts translating next to a start codon. The Kozak simply happens to be an efficient way to get a ribosome to start translating. The closer a sequence is to the consensus Kozak sequence, the more efficient it is, because the probability of translation initiation is higher.
