Weather balloons are inexpensive because the radiosonde payload is very cheap and light, not requiring much power or other infrastructure. Putting a proper surveillance payload on it would dramatically increase the price mostly because it'd then have to power the payload. That role is currently mostly done by UAVs.
Weirdly, I'm probably the one person who's worked extensively with both paradigms...
There are several players in the balloon market spinning up. It's becoming very hot very fast.
The balloons in question are much smarter that radiosondes now. They are fully capable of autonomous navigation within certain limits and can even orbit a point for days at a time via some clever weather mechanics. They're also very easy to launch. Two guys and a truck can launch over a dozen in a day and they're all aggregated and flown remotely via various links.
On the payload side it's actually much easier to do optics on a balloon, you have all the same pressure and thermal issues as space but you can iterate much faster when your cost to first pixels is thousands instead of tens of millions for the same ground resolution (GSD, drives necessary aperture size). I'm not going to spell out the details on either because that's the secret sauce that pays my mortgage but it suffices to say, if you're 10x lower, you need a lens that weights a little less. It's also worth noting that a VLEO sat will only be overhead for a few minutes, balloons can stream live video as long as you want, right now (no future tech or constellations needed).
The class of UAV that can begin to compete with a balloon is two to three orders of magnitude more expensive and still has nowhere near the endurance.
There are also other sensors and phenomenologies that are wildly more capable on a balloon platform than any type of satellite or UAV but I'll get yelled at if I spell any of those out...
There was recently an article about the 'UFOs' actually being 'just' Chinese balloons designed to record everything about the radars poking them to learn about capabilities of US defense systems (including F-35 in action). Guess we'll see more of them soon.
LEO is cheap but you need a massive aperture, measured in meters, to get equivalent GSD to even the crappiest balloon imager, so the price tag suddenly jumps. These aren't cubesats, they're suddenly the size of a bus, see Worldview Legion for a recent comparison, and it has a much worse nadir GSD than balloons.
VLEO is still extremely expensive because you need a very robust propulsion system and there are other design considerations like atomic oxygen corrosion. The optics to match a balloon also put you into the mini-fridge to refrigerator sized optics assembly class which means while you can rideshare, it's not cheap to build or launch, see Albedo space:
https://albedo.com/post/upcoming-launch-of-clarity-1-and-alb...
Also the regulatory issues are still massive, you need to get a NOAA license for imagery and once you go under a certain GSD limit they become very difficult to obtain.
The can sometimes leak slightly but they make up for it by dispensing small amounts of ballast material.
They're generally more environmentally friendly too, worst case you have to go ask a farmer to retrieve the thing out of their field.
Please tell me, in the industry, this is called a "rapid unscheduled descent".
If it isn't, it should be! ;)
_In principle_, a VLEO satellite could stay up forever (or, well, realistically until the engine wore out).
Weather balloons also don't last forever.