Why is my module interlocking mechanism failing when I scale it up?

The mistake alot of people make here is assuming clearance should scale with the model. It shouldn't. The gap you need is set by your printer and the material, not by how big the box is. A 0.2mm clearance that worked on your small prototype still works on the big one, because the printer's XY error is basically the same no matter what it's printing. So if you're multiplying your tolerance variable by your scale factor, stop doing that and keep it fixed. That alone probly fixes most of the weirdness you're seeing.

What I'd do in OpenSCAD is define the clearance as its own variable at the top, somthing like tol = 0.2, and then everywhere you cut a slot or fatten a tab, offset by tol explicitly instead of deriving it from the overall dimensions. That way scaling the box changes the outer size only and the joint fit stays constant. If you really want to scale everything at once with the transform, use the scale factor for geometry but pass the same tol into your modules, since scaling a boolean-cut part scales the gap too.

Also worth checking: is the looseness actually from the design, or from the print? Big flat parts in PLA can warp slightly at the corners and shrink a hair on cooling, and a long diagonal seam will show that way more than a little prototype did. If the first layer is squished at all, elephant's foot eats into your slots. A brim and a slightly higher first layer usually sorts it. For the joint itself, plain rectangular tabs will always have slop in one direction no matter what tolerance you pick. If you angle the engagement surfaces, even just a few degrees, so the tab wedges tighter as it seats, the same clearance suddenly feels snug because any play pulls it into the wedge instead of sliding along it. On my card box I did that and went from rattly to solid without touching the tolerance at all. Last tip: print a small joint test piece at each new size before committing to a full box. Ten minutes of filament saves a four hour print that doesn't snap together.

One thing I'd add is to caliper a printed test piece before trusting any of your numbers. My printer runs about 0.1mm wide on a 100mm move, which sounds like nothing untill you stack a tab and a slot and a hinge pin togather in one corner and suddenly the whole clamshell fights you. I keep a little gauntlet of single joints printed at each deck size and measure the actual slot width, not what OpenSCAD thinks it is. Feed that measured value back into your clearance variable and the scaled versions start behaving. Also, keep the whole joint as one module with the clearance as an arguement, so you're only ever tuning it in one place.

I've been messing around with some interlocking fixtures for a tool organizer, and I'm running into similar fit issues. It seems like the whole joint works okay if I print the parts flat on the bed, but if I try tilting them to minimize supports, the tabs just shear or drag way more than expected. I'm not really sure if that's just a basic layer line problem or if I'm messing up the mechanical strength of the joint itself when the parts aren't sitting perfectly horizontal. If the load is mostly shear, does the print orientation really matter that much for the tolerance gap, or is it just about getting a solid layer adhesion?