Here is a reference to a USGI test done in the 60's:
link Shows five years compressed decreased spring rate from around 5.5 to 4.6, but they functioned (m14) (and they had determined that 4.5 would be the lowest limit to "perform satisfactorily")
This shows some bad initial "set" but compared to a used spring not as bad as it looks in first pic:
link
That is the best I can find searching for now....
The first link conlcluded that storage loaded does NOT effect them, yet the spring rate in 5 yrs dropped from 5.5 to 4.6, and it says 4.5 is the lowest they can expect to function. I would guess that after 7?, 10? years the spring rate would drop down below the "functional" level of 4.5. That test was done in the 60's and newer metals in the springs may make that test worthless for today's mags....
ETA: interesting in the first link (Springfield Armory) all the mags were within functional range after 6500 cycles, and 2/3 in range after 10,000 cycles. They were cycling at almost 2 strokes/second! That makes we really wonder about the "it's loading and unloading that wears out the springs" internet "fact"....
They only had a sample of three in each test, so certainly lots of room for statistical error, but interesting.