It's good to know. But I don't have both of these cases. It doesn't make sense to store temporary data in QVD. And single parameter Exists is not possible because all possible values for the field already exist.
You should comment out the top portion and load just the first load statement and record the time. Then uncomment each preceeding load statement one at a time, and perform a reload to capture the time. Use the time differences to figure out where you really start to loose performance. I would also monitor PC and RAM and write down some notes each step of the way.
Also, as stated above its hard to really know the impact of a suggestion without using your hardware and data, but I would try the following:
1) Seperate Day, Date and Hour into different fields this will make the possible values less distinct and possibly decrease your memory footprint. Probably only a slight difference but it may add up for you.
2) All of your date functions result in Dual value outputs, try putting a Num() around them so only the numeric value is passed to the next load. Not sure if the overhead of the Num() will outweight the removal of the Daul value. Only testing will tell.
3) You may want to try using numbers as your IDs not codes. 1, 2, ... may perform better than EST, GMT, ... again I would test.
4) Try saving the Temp resident table to a QVD and dropping it, and then loading it back into memory with the where statement, this may sound crazy but this will cause QlikView to compress the data and free up the memory. Loading a large resident into another resident is ofter a bad idea because until you drop that Temp table (that won't be needed in the end) your using twice the memory.
I don't promise any of these suggestions will work, but I would try each one and record the results. Please share the results I would be interested to know whether each suggestion has a negative or positive impact. There are also other things to consider like what are your hardware and software specs. I have seen machines where the QlikView load process slows down if the CPU affinity is too high (this only started occuring after the multi-threaded loads were introduced).