I was always told to do them in parallel as is one blows then you don't loose the circuit,
Who told you this? The only way I can see blowing the whole circuit would be by using the wrong power source. In this case it would blow the circuit whether or not the data was wired in parallel.
It is possible to blow the first LED in the chain if your ground is disconnected. If the ground line becomes unattached, the LEDs will still work for a very short time. During this time the returning current from the LEDs will flow through the signal line of the NeoPixel. At least this is my present theory.
I had a hairbrained idea of trying to power some NeoPixels on my robot's wheels. My "power transfer" system apparently had some glitches.
The first pixel of my 16 pixel ring kept dying when using this setup. The OSHPark PCBs are good quality but after replacing the LED on the ring a few times, one of the traces was damaged.
Here's my "blue wire" repair.
So technically, your parallel wiring scheme places the first LED on each of your PCBs at risk. If you wire them in series only the LED with its data line connected to the Arduino would be at risk.
I bet the series resistor AdaFruit now recommends reduces the likelihood the first LED would be damaged. I'll probably start using a series resistor myself and see if I can reduce the abuse I give to my NeoPixels.
My power transfer setup is an extreme example. Under normal use, the NeoPixels should work fine.
and the capacitor is relevant and the correct uF?
I don't have have an answer for this. I think AdaFruit generally does a good job with their tutorials and I'm sure they have a lower kill ratio when using NeoPixels than myself. I don't think the exact value of the cap matters much. I'd think a few hundred uF would likely work okay.
My one concern with using a 1,000uF cap is the in rush current when power is turned on. I think a 1,000uF cap could damage a USB port if you tried to power your setup from the USB supply. I'm not sure about this myself.
As I mentioned earlier, I don't use a series resistor (though I'll likely start using them) and I don't use a 1,000uF cap in my NeoPixel projects.
I've had trouble when controlling NeoPixels from a 3.3V microcontroller. NeoPixels don't appear to like 3.3V logic when powered from 5V. You can see glitches in this video caused by using 3.3V logic.
Fortunately there are lots of logic level converter options and another option is to power the NeoPixels from 3.3V. Either of these solutions fixes the glitches.
Of course the Leonardo is a 5V controller so you don't have to worry about this. (So why did I tell you? Oh, that's right, so I could post the "hypno rings" video.)
Yes but they are not like Christmas tree lights. The light itself can fail but the electronics that pass on the signals often still works.
This hasn't been my experience. When I've had the first LED fail, it completely failed and didn't pass on the signal. Still this is actually an argument against using the boards in parallel since in series only a single LED is at risk of damage from a bad ground connection.