I love vague terms like "long periods". Long compared to the Planck length? Geological time? Is the advertised 43 seconds almost there or "off by 17 orders of magnitude?"
“Off by 17 orders of magnitude” would be off by 136 billion years, so not that much for sure. Assuming you want to be able to test the plant and or maintain it once per year, 43 seconds is less than 6 orders of magnitude off. The jump was more than a full order of magnitude compared to past records, so another handful such developments and we are there.
This is shared in the better article here: https://www.ipp.mpg.de/5532945/w7x
> During the record-setting experiment, about 90 frozen hydrogen pellets, each about a millimeter in size, were injected over 43 seconds, while powerful microwaves simultaneously heated the plasma. Precise coordination between heating and pellet injection was crucial to achieve the optimal balance between heating power and fuel supply.
Nuclear already gets taken offline for several weeks for refueling, but redundancy covers such issues.
Anyway, the older generation of devices was pulsed for engineering reasons (like non-superconducting coils getting too hot). The current generation of device is solving most of these and is limited by MHD instabilities alone (neoclassical tearing modes, mostly), if we can get active control mechanism working, then will be finally approach the long-pulse or steady-state regime.
A smooth toroidal magnetic field cannot confine plasma. The field at the outer side (further away from axis) are spread more widely and weaker than in the inner side. In a very short time, this will cause ions to drift out of confinement at the outer side. The solution is to produce a twisted, helical field, where the field lines go in circles in both directions of the toroid simultaneously, like the stripes of a candy cane in the bend.
Different reactor designs have different solutions to this. Tokamaks use a solenoid to drive a strong toroidal current in the plasma. This, in turn, causes a poloidal magnetic field, which provides the second half of the field needed for confinement. But this only works when magnetic field of the solenoid coil is varying smoothly over time in a single direction. Eventually, you hit some limit in your ability to do that, at which point you lose your ability to confine the plasma and the pulse ends.
Stellarators do not have this issue. They get the full field geometry needed from their primary field, by twisting it around the toroid in a very complex path. The downside is that they are much more difficult to design and build.
This is a much less sexy problem than containment, but it's a showstopper for commercialisation. You can just about imagine an epically huge reactor with unfeasibly powerful containment fields that trap fusion in the centre of a large cloud of hydrogen, which captures neutrons to make tritium to power the reaction. But that's completely unbuildable with current tech.
Aneutronic fusion is possible, but it happens at even more extreme temperatures, which are barely theoretical at the moment.
At this point we've been chasing fusion for more than 70 years, and commercialisation is as far away as ever.
You might as well just build yourself a small star.
Or perhaps even spend all that research money on making better use of the star we already have.