The way things should work is like this: Subscription for when you want to use the software and open format data files that you're always easily able to export and download.
If a company goes out of business or does something stupid then another one will pop up to handle your needs. This aligns incentives correctly while still allows a healthy dose of competition.
The "middle ground" has been the most common model for desktop app pricing for decades. You buy version X of a product. When the next major version is released you pay an upgrade price.
In fact, the "middle ground" has been the norm for the majority of software apps over the history of desktop computing. It's the SaaS subscription model for cloud-hosted software that is the new deviation. But that cloud-based subscription model is now leaking into desktop apps (like the Adobe suite).
Mobile app stores killed that model, when they made continuous, perpetual updates the norm. At the time, everybody was too busy pointing out the implications of going from 50$ to 50ct when the even bigger change was a farewell to major version upgrade sales.
But there, after a few cycles of companies ruining themselves by hiring in the short-lived money surge of growing an app to saturation, parts of the industry seem to have found their way to quite reasonably priced subscriptions.
Continuous, perpetual updates are definitely NOT the norm in mobile apps.
Some get some bug-fixes and the occasional new feature, but any major update lands a version 2.0.
Far more (the "long tail") are just abandoned (like all those thousands of 32-bit apps being deprecated in the iOS store).
2015 - you sold v1 for $5.00
2017 - you start selling v2.
You sell v1 for $5.00 but in the description tell everyone not to buy it.
You sell v2 for $5.00 but direct people who have already bought v1 to buy the bundle of v1 and v2 that you price at $8. Anyone who has already bought v1 can buy the bundle for $3.
N * (N + 1) / 2
Not very satisfactory, but very prevalent.
And, I'd say, given the low price of apps, it might be just right to not offer any special discount for an upgrade. You could always reward older users some other way.
Also, Adobe's tools have always been among the most pirated software out there. People who needed to do occasional work wanted Photoshop but there was just no way they were going to drop $700 on it. Nowadays, Adobe just charges them an $11/mo fee. That makes it much easier and I know several instances where this newer model has resulted in real income for Adobe, from what would've otherwise been pirate installs.
People are much more willing to pay a small fee several times than a massive fee once. The subscription makes them feel like they have more control and they can cancel when they don't need the product anymore (not always true because some of Adobe's subscriptions have annual lock-ins). It reduces the friction from "figure out how to pay us more than your car payment" to "just send us the cost of a trip to Starbucks, it will only take a couple of minutes and you'll be merrily on your way".
This dramatically changes the incentives re: piracy. Instead of spending hours trying to find a cracked build from a site that won't pwn your PC so that you can save hundreds of dollars, you just pay a few bucks and speed through with a known-good version (plus a clean conscience).
It's almost a side benefit that subscriptions frequently represent additional lifetime value v. a "pay once" model.
A user can buy the base version, then a few months later when you create a release that allows processing widgets through the program twice as fast, they can update for a small fee. This would force the developer to create updates that would be easy to communicate the benefits of to the consumer.
It'd be a bit more difficult to figure out pricing for new users over time and which version they receive, but it would allow they old pricing model without forgoing the benefits of incremental development over time.
That could quickly devolve into supporting a large number of distinct versions of the software, which would be difficult and costly. Nobody wants to do that.
Not to mention that, if this versioned software is a platform (for example, WinXP and its service packs you mention), downstream vendors would also need to QA and support their software on all actively supported versions of said platform.
I use various bits of successful software like Transmit and Alfred that let you update until the next major version in which you must pay again.
The reality is that having everyone on the same version drastically simplifies not just security, but also interoperability. It isn't 1998 anymore. The default is internet connectivity and we should stop pretending our apps are secure when they are not.
That being said I do think truly securing an application requires first having a secure platform. Otherwise users could indefinitely demand mitigations at the application layer until it contains most of its own OS.
Now you do you, and I'll use Goldwave v4.0 for basic audio edits - exactly the same tool I used in 2000, one that works well and I never had to re-learn.
I despise tools that capriciously change in my hands. Extra "security" is not worth the productivity hit.
Again, Internet connectivity isn't my default. Dropbox can handle that, and that software gets all the updates. The tools I use to make stuff I put in Dropbox better all be offline and never changing unless I say so.
That applies to the office suite, DAW, graphics editor, etc. I haven't seen anything their online features could give except a cloud lock-in.
You may value extra security, often necessitated by the very virtue of the software being online. And I value the ability to use without having to think about whether the signal is strong and my card payments went through.
The reality is that the interests of the users are often not aligned with the interests of software vendors.