I think so too; I went ahead and tested it for myself...

I think I got pretty similar results. So probably no new entities to create this effect, just good 'ol blocklight brushes and high intensity lightmaps...

Well, I went ahead and conducted a test for this on my older machine that has these specs:

Model: Dell Dimension 4600
CPU: Pentium 4 @ 2.7 GHz
GPU: Nvidia GeForce FX 5200 with 128 MGb
Resolution: 1280x1024 with 32 bit color depth

I made 4 different maps each testing different setups for the shadows:

"test_lightmap_grid" =LG (This means I set the map up with a lightmap scale of 1 on the surfaces producing the complex shadows and included the "blocklight" grid of brushes to mimic the grate's shadows.)
"test_lightmap_NOgrid" =LNG (This means I kept the lightmap scale of 1 but removed the "blocklight" brush grid from the map meaning no cool shadows.)
"test_NOlightmap_grid" =NLG (This means I set ALL lightmaps to a scale factor of 16, but kept the complex "blocklight" brush grid.)
"test_NOlightmap_NOgrid" =NLNG (Meaning I set the lightmap scale to 16 AND removed the "blocklight" brush grid. a.k.a. just a blank map with a light.)

================================================================
In the first test I set all the settings in Half-Life 2 on high except for Vsync (which I turned off) and the filter mode to "Trilinear." Here's what I got:

(Above) This was map: LG which gave me 62 fps.

(Above) This was map LNG which gave me 63 fps (just one higher for no shadows).

(Ab..o...v...you get the idea!) This map was NLG which gave me a considerable jump of 68 fps.

And this was map NLNG which gave me just one more frame of 69 fps for no shadows.
================================================================
Now, I conducted one more test with the maps, but this time I kept all the settings in Half-Life 2 the same except for the Filter Mode which I ramped up to: anisotropic 16x, and here is what I got:

This is map LG, which now having anisotropic filtering of 16x gave me a considerable reduction of seven frames giving me 55 fps.

This is map LNG which gave me the same thing at 55 fps.

This is map NLG which just gave me one jump of 59 fps.

This is map NLNG which gave me a jump of 58 fps (it was jumping between 58 and 59, so really no difference)
================================================================
So here are the results in summary:

Test with Trilinear filtering (in fps):
LG=62
LNG=63
NLG=68
NLNG=69

Test with anisotropic 16x:
LG=55
LNG=55
NLG=58-59
NLNG=58-59

Conclusion:
I don't believe it is the shadows that slow down system performance, but rather, the lightmap scale that does so. So, if your deciding to give a surface a lower lightmap scale, then by all means, take advantage of applying as many cool looking shadows to it; it will just be affected by the lightmap scale and not by the number of complex shadows painted onto it...

Ah, good point, here are the file sizes with the same code for the map names:

LG=1.09Mb
LNG=1.02Mb
NLG=146Kb
NLNG=76.8Kb

(It was just that one room)

Well, Reaper, you're right, the shadows do affect file size, (and of course so do lightmaps) but if you're going to add low scale lightmaps to your level, might as well make 'em look good.