What you are suggesting is common practice in the billboard / very-large-format-printing industry.
First, go to:
http://graphicdesign.stackexchange....on-should-a-large-format-artwork-for-print-be
Scroll about 40% down the page to the section titled, "Raster Image Data" and read it.
Basically, in contrast to some of the other posts in that thread, this section (#11) is written by a knowledgeable person, and says that for printing large items, you rarely need the high ppi numbers that you might think you need, and often, surprisingly low ppi numbers are entirely adequate, and even preferred by the printing company.
Then, and only IF you have established that you actually need the 150 ppi that you stated in your first post, and can't work at 50 ppi (ie, with smaller pixel dimensions), there is a technique called, "scaling" that is also very commonly used in the billboard industry. It is done exactly for the reason you gave.
The basic idea behind this technique is to do all of the development and refining of the adjustments that your image may need on a low resolution (ie, low pixel count) surrogate version of the image, and then, if the adjustments are mostly adjustment layers, you can just pull the layers over from the surrogate to the unwieldy, full resolution version with the high pixel dimensions. You bring all the adjustment layers over, and let PS chew through them only once. This way, you are not waiting 4 minutes every time you want to do an experiment with a slightly different adjustment.
The basic idea that I described in the previous paragraph runs into problems if your work on the image includes sharp edged, detailed layer masks, compositing, or local (not global) adjustments. If you need to do such things to achieve final product, you may have to do things like develop your layer masks or component images on full rez versions of those images, and bring them into your final PS file only once, near the very end of the process. The speed advantage of generating detailed masks and component images separately comes from the fact that you can do such work in one or more separate files, each with the absolute minimum number of layers, maybe as few as one layer, so you are taxing your computer resources vastly less than trying to do something like develop a detailed mask in a file that might have dozens of full rez layers.
HTH,
Tom M