Do I understand correctly that both send tracks should be panned center?
Shoot, maybe I edited the post too late. 
Anyway, no. One send to the left, and one to the right.
Added examples in the original post. Where do you guys usually upload the XRNS files?
You could use one of the many file hosts out there. Sendspace seems to be quite a popular one.
You could also use Google Docs if you have a Gmail account or something like that.
Personally, I think DropBox is great. You can get a quick tour of it here.
You get a nice amount of space on the free account (2GB currently, I believe), and it’s very easy to put files into your public folder and then share the link on forums and stuff. You can optionally install their software which gives you an auto-synced folder on your computer that is connected to your account, but you can also simply use the website to upload and manage your files.
If you like the look of it, you can sign up using my referral link below and we both get an extra 250mb free 
Done! Added the link to .XRNS file to original post.
Interesting stuff!
So… I don’t get it why you need to use send tracks for it…
Here’s the same thing without sends.
Also in your example you have the synth1 panned almost fully left so that only a little amount of the signal passes to the right channel. Muting the Pull Left send reveals that there is almost no ‘Björkifying’ effect.
That’s a pretty nice tutorial. 
Though there is a tiny “error” in it. Tiny, but with a huge impact.
To get the maximum effect out of this setup, you have to mute the source in the Delay DSP. This increases the FX amount of the original tutorial by 100%. And it sounds like there’s worlds between both setups. By not muting it, you kill half of the FX amount.
Damn, I was almost sure I tried it before, but obviously haven’t… Thanks for point that out.
You mean you can’t hear the difference between left send muted and unmuted? I just rechecked the file I uploaded to dropbox, and I can hear the not-so-insignificant difference. If I solo the track, disable all DSPs, and then monitor the master spectrum, I see both L and R have signal…
My first thought was also, a Delay with a muted source should do the same job. But in fact it doesn’t.
What happens in the tutorial setup is, it doesn’t just delay one side. It delays one side and adds the delayed signal to the non delayed signal, of the same side. The psycho-acoustic effect makes a huge difference. You can’t achive this with one delay only.
It may be that the synth1 preset sounds different over here. Not a good idea anyway to use VST-s in example .xrns files because some people don’t have them available.
Actually, eeter is correct. I’ve tested both methods, and there’s absolutely no difference (that I can hear). I added the instructions for eeter’s setup to the original post. Thanks, eeter!
Well, I can hear it. 
And I can measure it. It’s really not the same.
Give me a few minutes. I’ll put up an example to download with several comparisons.
You’re right. Compared them and there’s a quite big difference actually.
But I really don’t get it where the difference comes from.
You’re right about leaving unsampled VST. I added links to presets to the original post. Would the right way to post this be to export the whole track without DSPs as a WAV file? Plugin grabber totally didn’t do the trick (or I suck at grabbing plugins…).
That’s simple. Just follow the signal flow and think over what it does.
1st send + 1 receive = doubled volume on the non-affected side
2nd send + 1 receive (delayed) = summary of 2 time-displaced signals
So the tutorial setup layers 4 streams. 3 originally timed and 1 delayed.
With the common delay in comparison you just get one non-delayed side and one delayed. Which is only 2 streams, one of them delayed.
Btw… did you guys actually see me previous advise?
Try this. So the effect becomes REALLY obvious. Sounds REALLY wide.
Ok, I think I know why I couldn’t hear the difference. I have -3dB volume on both sends, so I don’t have that volume boost in the example. It’s either that, or my headphones need a replacement. I wanted to demonstrate the “expansion” part without making it louder than the original.
Ha! Awesome. It really makes a huge difference. Thanks. Added to the original post.
Uhm, wait… I’m telling crap. Totally crap. At least for the signal summaries, because these are not layered. The second send DSP is muted.
So, like eeter said, the output should indeed be the same, using a single delay. But it’s still a fact the setup’s output differs from the plain setup with a delay only. That’s pretty strange. I’ll try to get into this…
Anyway, here’s for a first step the corrected (muted resource) download.
Btw, in that test file, you delayed different channels in the plain-delay and tutorial-delay versions. It doesn’t change the fact they are different, but it might help get more correct findings.
EDIT: I’ve flipped the channels on the plain-delay delay DSP (left 40ms, right 1ms), and it no longer sounds different, just quieter.
LOL, okay. No need to wonder why the measured stereo spread differs then.
Anyway, this setup isn’t useless. When you add a Gainer on one of the send tracks and inverse its output the shared mono frequencies of both channels are eleminated on the summary. So you can still maximze the stereo feeling of the sound this way. Other side of the medal: this also might reduce frequencies you might not want to lose.