Well when the "arguments" against generally consist of "stupid design that'll kill people" or draw parallels with completely different designs (eg in Luton), it's easy to draw the conclusion that the people making them have never SEEN it work and don't really understand what they are looking at.
and...
err yes the 90% approach to the crossing is the whole point. If they don't that, it's dangerous. It MUST be the case in real life. This design isn't just "put a track around the outside" it's a basket of measures, all of which are necessary, the most important of which are the right-angle exits to slow speeds and massively increase visibility, and having enough space to stop.
Notice also that each arm has a different treatment in this test roundabout. They are playing with different options.
I'm also nervous about drivers in the UK and their ability to cope with this, but we have to start improvements somewhere. I realise that people in this forum think this is a backwards step, but well luckily even TFL is starting to disagree with you...
With respect, you're missing the point. In the example with the arms at 90°, the video on the news clearly showed a car stopping on a roundabout to give way to a cycllist, whilst on the other side, a car went merrily across and would have hit a cyclist, unless the cyclist hadn't stopped. Whilst I agree entirely it would be great to completely re-engineer car driver's attitudes, it's going to take years. To have a car suddenly stop on a roundabout, will cause accidents, and cyclists will be injured and killed if this sort of roundabout is introduced.
Having been to TRL last week, even though we didn't get to go on the magic roundabout, after talking to the staff quite a bit, although they have good intentions, I'm not sure they really understand all the issues faced by cyclists in a busy urban environment such as London. For example on the traffic light junction we were testing, they were trying different combinations of alternate working lights and timings. However most of the time it was with just 1 cyclist at a time, and there were no cars at all. So I did query with them how valid their data would be, and they did agree, but said they tried using cars as well, but found the tests took far longer.
Edit: Also, I should point out contrary to what you've said, the Luton roundabout isn't a completely different design. The TRL example have raised divisions, whilst the one here in Luton uses tarmac kerbs. In both cases cyclists are segregated. The TRL one has cyclists with the right of way all the way round. In Luton, cyclists are expected to give way at some of the interactions, but not all.