Imagine you are a football player (the processor) and I am your coach (the OCer).
Say I tell you to run down the field and back and you can do it in 1 minute.
Now pretend I want to overclock you and I tell you to make the same run except this time in 55 seconds. You now have to do the same amount of real steps (cycles) in a smaller amount of real time.
Alternatively I could ask you to do more steps in the same amount of real time, like if I told you not to just go to the goal line, but all the way to the edge of the field, adding a couple more meters each way and I told you to do it in the same 60 seconds.
Both are different ways of describing the same thing.
Say I see you can indeed make it to the end zone and back in 55 seconds so this time I tell you to do it in 50 seconds (a more aggressive OC). Pretend you do indeed make it down and back in 50 seconds (a stable aggressive OC).
Say I ask you to make it in 45 now.
Eventually, one of these times you just physically aren't going to make it unless you sprint as hard as you can the entire time.
Pretend that your blind sprint speed is 45 seconds and you really can't drop even 1 more second off the time, that would be your ultra aggressive OC time.
The way OCers generaly operate, they will just leave it at that setting forever.
It would be as if I told you to just sprint as hard as you can indefinitely.
What is going to happen during this process is that eventually you will 1) trip and fall on your face and have to get up and resume sprinting (the cache miss), or 2) you will get winded and slow down (throttling yourself).
A processor cache is basically RAM that is physically on the processor. It is only a tiny amount, only a couple megs at the most. Data is put there by the processor when the processor wants to send it to the RAM.
In terms of how fast a processor operates, sending something to the RAM and receiving something from the RAM are extremely slow. It is only a couple micro seconds in real time which is meaningless to a human, but it is a really boring wait for a processor.
A cache hit is basically when the processor stores something in the cache to be sent to RAM and then it tries to access the same information again the tiniest split second later and the processor finds that the information is still waiting to be sent to the RAM, IE its still in the processor's built in RAM.
When this happens, the processor can delay the send, change the information in the cache, and then send it with the updated information.
A cache miss would be if the processor went back for the information and found out that it just got sent to the RAM, now it has to send out a request to the RAM and wait for the response to come back so it knows what the data is.
In people terms you might describe this as if you wake up and remember that you need to mail in your power bill so you jump out of bed, write the check, stuff it in an envelope, and you start sprinting for the mailbox.
A cache hit in this scenario would be the mailman hasn't come yet today. A cache miss would be that the mailman already came and he won't be back for more mail from you till tomorrow.
Basically, the faster and faster you OC a processor, the more often you hit problems like the cache miss. The more you speed the whole thing up, the more often that the cache has to be written to the RAM and the less time stuff stays in cache.
Essentially, as you get more aggressive and more aggressive with the OC, the number of tiny slip ups the processor makes increases just like the likelihood that you trip and fall on your face increases the faster you are made to run.
Contrast with how most people pretty much never fall on their face when they are walking at regular speed on a mostly flat surface like most sidewalks.