My initial reaction is to say that computing is simply cyclical; what was once mainframes and dumb terminals turned into locally installed applications on desktops and laptops, and now we're doing that again with Teh Cloud (tm). However, here's the difference:
1.) In the 80's and early 90's, overall technical competence of computer users was higher. Yes, the there was always the secretary who tried to use WordPerfect to make a database because she knew exactly one program, but overall, especially if you had a home computer, you had some concept of what you were buying, and what the things on the spec sheet meant - computers being sold today will have helpfully descriptive bullet points like "great for multitasking" instead of "8GB RAM", something that wouldn't have passed muster in the last cycle.
1b.) Malware was much less a problem, back in the earlier days of computing. E-mail viruses were a thing, certainly, but for the most part, one ran a virus scanner and moved on with life. Also, with less hardware to throw at resident software, any kind of malware that ran resident would use enough system resources to alert the user to its presence, which is less the case now. Google Docs doesn't care about macro viruses, and users of that platform don't have to, either. There's value in that proposition for many less-technically-inclined users. Similarly, backups/hard disk crashes are "someone else's problem".
2.) In the 80's and 90's, systems were generally designed for interoperability a bit better than they are today. It's possible to send an e-mail from a server running Exchange 2016 Preview to an SMTP server from 1989 and it'll be able to meaningfully use the message. This is not the case with Facebook or WhatsApp.
3.) Inherently connected applications are the norm now. The utility of Facebook is "the rest of the stuff on Facebook". Google Docs and Pixlr don't apply to this point since they still deal with .doc and .jpg files that are more standards compliant, but many of the web apps that are popular aren't necessarily tied to the "open/change/save/close" paradigm that is commonplace in the desktop world.
4.) "Bleed little, bleed often" is a more culturally acceptable proposition for most people, as it gives them the instant gratification of getting the product at a price they can afford, while not requiring a gargantuan up-front cost that happens regularly as people feel the need to keep up with the Joneses. $5/month = $180 over the course of three years, which has basically been the shelf life of every version of MS Office released. Makes it a lot easier to swallow for many people, whether or not it's actually a value proposition in the long run.
4b.) The fact that virtually every software developer who has implemented IAPs instead of a one-time, up-front cost has made more money on that business model. At this point, it's solely a matter of principle that a developer of a paid application would sell a perpetual license, since general acceptance of subscription and IAP licensing makes it a better idea for everyone to go down that road instead. This was not nearly as true in the days of mainframe computing.
Now all of that being said, I do think that video editing is one of the few tasks that will never lend itself to a subscription model, beyond what Digital Juice does. Editing-as-a-service makes very little sense, since even a moderately sized project will still take tens of gigabytes of upload time, which means "hours before you can edit". Meanwhile, 100GB of assets is not unheard of for even a two hour wedding video shot in HD, and with upload speeds still measured the single-digit mbit/sec unit, it can easily be days before editing can even be entertained. At the same time, costs are a lot higher for a company looking to get into that business, because you're going to get much less ability to thin provision even 500GB of space, as the nature of what's being done is going to make much more use of that space than the OneDrive accounts with a 1TB progress bar. CPU/GPU processing is similarly much more likely to be leveraged, and we haven't even gotten to the headache that is "plug-ins" - Alien Skin, NewBlue, Red Giant, Pixelan, BorisFX, Panopticum...and that's just the list off the top of my head of companies that make plugins for video editing platforms. How, exactly, does that work in the cloud? Now you're talking about a sub-subscription, or deals whereby plugin developer FOO has to pay Adobe to be included, and the number of subscribers is higher than plugin developer BAR, so they pay less per seat, and it's Net Neutrality all over again. All in all, to create a viable video editing as a service platform, it'd be about the cost of a new, moderately-well-spec'd laptop per year in order to make a profit off that particular niche...and if that's the cost, business pragmatism makes it very difficult to justify subscribing in the first place.
Meh, what do I know.