I think the framing of what's happened so far as "failure" is probably the main thing responsible for this perception. It's true that progress has been slower than many had hoped and the most optimistic had projected, but "failure" sort suggests that the things the research community have been trying haven't represented meaningful progress towards the goal of power production, which isn't the case.
Q (the ratio of energy out to energy in) has improved by about four orders of magnitude since controlled fusion was first achieved, and it's been a slow, at least reasonably steady march since the middle of the 20th century to achieve that progress. The current record-holding Q for magnetic confinement is around 0.67, so we need well under one more order of magnitude to get to the point of "theoretical break-even" (Q>1) -- we're most of the way there. A plant just barely better than break-even probably wouldn't be commercially viable, though, and while estimates vary, that point is probably somewhere in the 10-30 range, so we have maybe another order of magnitude to go after break-even. I don't think there's anything to suggest that after decades of progress we'll suddenly stop being able to make more.
It's true that things have slowed down somewhat in the last 10-15 years, but most of the blame there goes to the need, in order to continue moving forward, to build bigger and bigger reactors, and the need to divert resources to that goal (mostly ITER). To the extent that promises of going faster have turned out to be hot air, it seems like they've mostly been in the form of novel approaches that do fusion in some fundamental new way that avoids the need to build an ITER-like thing. These approaches seem to often involve lots of unknowns, and end up getting bogged down in practical issues once they're actually tried (surprise plasma instabilities and so on).
Recent advances in materials science (mostly REBCO magnets) and computing, though, offer a path to progress on the regular, bog-standard flavor of magnetic confinement fusion (tokamaks) on a smaller scale -- that's what this is. The nice thing about that is that the plasma physics here are very well understood, and have been heavily researched using conventional/not-super-conducting magnets that won't ever achieve break-even, but create identical plasma conditions inside the reactor (MIT Alcator C-Mod is effectively the conventional-magnet predecessor to this project). Up until now, the only real question was whether or not they could build strong-enough REBCO magnets, and now they have, so this is all good news and reason for optimism.
Of course, commercial viability is a whole other question involving lots of questions besides physics. But the physics here seem to not be in serious doubt, unlike some of the proposals from other startups that are more exotic.