What makes science fiction different from most other kind of fiction is that the author has more "overhead"--the explanations needed to make sense of what's different from the present day. A story has to have characters, plot, ideas, and overhead to work. Golden Age SF shorts made this work by sacrificing characters and sometimes plot to fit the overhead and ideas into their word count.
One shortcut for this is "used furniture." A story can transplant another setting into the future so readers can quickly translate everything. So "space western" stories have more room left for characters and plot by tossing out overhead explanations. Fantasy writers use the European middle ages for the same reason. Sometimes this can save more work for the writer than the reader--Weber's Napoleonic Age references in the Honorverse
The complaints about new short SF stories is that the writers are relying on a different set of used furniture--the past decade or two of literary SF ideas. That's great for readers who've kept up, they can follow the story just fine. But anyone coming from the outside--say a young Babylon 5 fan looking for something to read--gets confused, bored, and looks for something else that's more fun.
SF novels are doing better than shorts because they have room for all elements of the story on top of the overhead. Series are popular because you get more story pay-off for the reader's investment in learning the overhead. But a stand-alone short story can't do that.
The short stories I've read recently are spin-offs from longer pieces of fiction. Bujold has a couple of shorts in the Vorkosiverse, naominovik has one up on her website. They work well because almost all the overhead is in another book. Fanfic works the same way, capitalizing on the parent work's investment. So that's one model for successful short fiction. This is analogous to "DVD extras", things that can't stand on their own but the fans still seek out. Baen has sold story collections on that model as Honor Harrington spin-offs as well as the "Grantville Gazette" collections in the 1632-verse. Lee and Miller have sold shorts in their Liaden universe as chapbooks.
In the heyday of Analog many authors had series of short stories that would build on each other. That only works if you can depend on the readers to have read the original stories, otherwise each one has to have enough overhead to stand on its own. Or--and this is working for webcomics such as Schlock Mercenary--you could make the earlier episodes available online. That lets readers immediately answer any questions they have by going back to the earlier story with that bit of overhead explanation. Putting up shorts online for free horrifies the writers who think every bit of writing should be paid for in advance, but a few pioneering webcartoonists have made successful livings from the stories they've put out for free.
A third model for shorts might be the "free sample". An author with a book on the market could post a short story in the same setting--not an excerpt--as a way to intrigue new readers who might buy the book. This could be a good home for those lovely scenes that were fun to write but had to be cut to keep the plot from bogging down. Technically these may not be "short stories" since they still can't function on their own, they depend on the reader seeking out more of setting to be a satisfying experience.
I think the stand-alone short story is a relic of the 20th century. Modern readers want a full story with a plot and characters they care about (whether or not they're textured enough to satisfy critics). The magazine format can't work because it disconnects the story from the context needed to enjoy it. An online magazine could work if it presents every story with links to its setting (the archive of other stories and possibly a faq). Like webcomics, a webzine would need to build up a critical mass of material before finding people willing to pay for it. Trying to convince new readers to pay for shorts in advance is a futile exercise.