Right... The water temp within the loop will be very similar throughout the loop because the water is travelling so fast that it doesn't dwell within a block or rad long enough to change temperature IN COMPARISON to the water immediately after the block or rad... However, that doesnt mean that the water within the loop isn't being heated or cooled by the rads or blocks.
When your PC is off, the water will equalise temp with the room its in. As soon as you turn on the PC the water starts taking on heat from the blocks, it also transfers that heat to the rads which in turn shed that heat to the air in the room. There has to be a difference in temp between the water and the air for the heat to transfer.
The reason we don't all run a single 120mm rad is because a single 120mm can only shed about 100-200w of heat before the water temp starts shooting up past where you would want it to (its ability to shift heat actually gets better the hotter the water is, the bigger the difference between air and water)... The water in the loop would still all be the same temp though, just hotter than you would like.
If it helps to visualise it, the water just after the gpu will be say 0.5C warmer than the water just after the rad... But the water in the loop will be 5-10C warmer than the air in the room... That is where the heat from the PC is going, in to the water, that heat goes round and round the loop and some of it leaks back out of the radiator to the air, the more rad space you have the more heat is shed to air and the cooler the water is is comparison with the air/room temp... If you have enough rad space then the water stops heating up past a certain point, hence equilibrium... If you have a cpu and a gpu and just a 120mm then the water keeps getting warmer and warmer until a either the computer crashes or tubing/fitting/pump fails
Basically you wouldn't want 60C water in your loop
GPU blocks are less flow dependent than CPU blocks as they have a much larger surface area, parallel lowers restriction so putting gpu blocks in parallel helps flow rate in the rest of the loop, though actually again, with gpu blocks being low restriction it is pretty minimal... You absolutely don't want to put a gpu and cpu in parallel as the cpu block being much higher restriction will really suffer flow rate if you try
As mentioned above you would have to increase the restriction of the GPU block, which would also increase restriction in the whole loop, so again you'll be worse off than if you just ran them in series, parallel also halves the flow rate in each block, not a problem for a gpu block, but not ideal for a cpu block that relies on fast flow rates for efficiency
Watercooled GPU's usually sit well below 50C, so your CPU is what you want to concentrate on... Putting your CPU in parallel with anything else will halve the flow rate it gets, which is never going to be good
