> Won´t it have any DC running through the interstage windings?
Yes. That is why you use a medium-Z winding: they take more DC than a winding with more turns.
> Wouldn´t the tubes be "seeing" a 1k load with it?
Yes. The tubes and load need to be NOT "matched". If load and plate resistance are similar, then changing tube current has little effect on gain. The load must be greater than the maximum (low-gain) plate resistance, or less than the minimum (hi-gain) plate resistance. In this case, either over 300KΩ, or under about 2KΩ. Fairchild went hi-Z by using a bunch of tubes to lower the plate resistance (and still it needs a pretty fancy winding). I went low-Z for low-price: $100 versus $10,000. The disadvantage of the low-Z loading is low tube gain, but gain is cheap today: that 5532 output stage costs $2.
If you have a healthy 600:600 or 2K:2K, use that, don't spend the $18 for a pair of RS-brand cores. OTOH, if you don't have a pair of $95 trannies handy, try the $18 dual-trans gizmo first. I have used it a LOT to break ground-loops, and never noticed much "color".
In any limiter building, you need a good signal generator and good AC voltmeter. Levels through the tube-stage are moderately critical. Comments like "good for light compression" may be inherent in using a 12AU7, or may be levels very different from I simulated. (And of course "light compression" is a matter of personal taste: I whacked those organ-pipes 2dB and wished I hadn't; in other works I've slapped-down a runaway singer by 15dB and maybe could have hit her harder.)
Have I built it? My building these days is limited by time, eyesight, and especially now by back-spasms. I'm going to need a walk around the building after this short note. No way I can hunch over a solder-iron this month. But Kent and others have the enthusiasm I once had, and I'm trying to trick you into doing my solder-work for me.