If you search this forum, and the 'Net you will find a ton of info.
I claim "some passing knowledge" as a retired mechanical engineer with 31 years at a "major" automobile manufacturer.
First, we must assume that the magazine manufacturers have "done their homework" and selected the "proper steel" for the springs.
In the real world, it is always possible to screw things up. Sometimes technical ignorance, sometimes simply "cost".
Given proper steel, and proper manufacturing techniques, it would be VERY hard to reach the thousands (hundreds of thousands)
of cycles needed to achieve "spring fatigue" to failure. And, leaving the spring compressed will NOT affect its life cycle to fatigue.
Given that, it is possible to screw up design and material parameters through "ignorance" and have springs which quickly "take a set".
Shouldn't happen, but it can. See above for "technical ignorance" and "cost". And "human involvement".
For these same reasons, it is possible to "make" springs in which a significant number will "fail early". Often this is VERY early.
There is a whole area of "engineering" in which testing and quality control measures verify "proper life cycle design" for springs.
For an example of what properly designed springs do millions of times a day, think of automobile engines and their "valve springs".
I will leave it to "the student" to determine how many times a single valve spring cycles during a theoretical 100,000 mile vehicle life.
Hint: a typical engine spends a lot of time between 1,500 and 3,000 revolutions per minute (RPM). And a lot of "minutes" in 100,000 miles.
Hint: while the vehicle is just "sitting around" waiting for the owner to use it, some valve springs are compressed, and some are not.
Or, for an at home example, my Beretta 92 was purchased in 1992. Both magazines have been loaded all the time since then.
Well, except when being fired dry, and then reloaded. Never a failure to feed in thousands of rounds.
I guess Beretta has "learned some things" since they started firearm manufacture back in 1550 or so.
As always, YMMV.