4 hands Siteswaps
This notation is used in certain specific cases:

It only describes patterns for 2 jugglers (which is where we get 4 hands), referred to here as J1 and J2.
 The patterns in question must be asynchronous. As per the explanation of causal diagrams page, these are patterns with a delay of 0.5 (with passes like 3.5p, 4.5p... in normal siteswap). This means that J1 and J2 never throw at the same time.
Since J1 and J2 don't throw at the same time, one might think of it as nothing more than a single imaginary juggler with 4 hands throwing the clubs one after another in the following order: RHJ1 (J1's right hand), RHJ2, LHJ1, LHJ2.
Thus we can assign numbers to each throw as we did in normal siteswap and obtain the following table showing the correspondence between the two:
4handed siteswap

Description

Normal siteswap equivalent

0

empty hand

0

1

impossible

0.5p

2

handacross

1

3

impossible

1.5p

4

pause 
2

5

almost impossiblevery
fast pass

2.5p

6

normal self

3

7

lofty single pass

3.5p

8

double self (straight across)

4

9

lofty double pass

4.5p

10

triple self (crossing)

5

11

lofty triple pass

5.5p

Remarks :
 All normal siteswaps are valid 4hands siteswaps. And that was actually the primary goal of this notation: finding an interpretation for passing of nomal one person siteswaps. You can take a siteswap generator, tell him not to use 1, 3 or 5's, and generate all possible asynchronous rhythms for 5 to 9 clubs.
 Divide by 2 to obtain the equivalent in normal siteswap.
 All even numbers are selfs, while odd numbers are passes.
 You will notice that I did not specify whether the passes are crossed or not. That's OK if J1 does 7 and 11 crossed, then J2 does 7 and 9 straight across, with 9 being straight for J1 and crossed for J2.
 I marked certain passes as being impossible. This is true in a normal configuration, but it could, however, be possible to do them as a handacross in backtoback, for example (for 3 and 5).
 Note that if 10 can easily be confused with a 1 followed by a 0, 1's are almost nonexistent in 4handed siteswaps. So 10, 11, 12, etc. are always read as ten, eleven, twelve, etc. in this type of siteswap, unless specified explicitly. They are sometimes replaced by a,b,c...
 You can find some more advanced stuff on the subject in Norihide Tokushige's
article: passing siteswap.
How to use it and examples
When faced with a 4handed siteswap, first we have to know to whom the sequence appliesthe 4handed juggler, J1, J2?
Normally, there's a sequence for the virtual juggler: S1 S2 S
3 S4 S5
...
and it is specified: where J1 does S1
S 3 S5
... and J2 does S2 S4
...
You can draw a table to associate each number with the hands of each juggler if you still need to convince yourself:
S1

S2

S 3

S4

S5

...

RHJ1

RHJ2

LHJ1

LHJ2

RHJ1

...

example 1 : 966 (3count with 7 clubs)
The 4handed siteswap is 9 6 6 9 6 6 9 6 6 ....J1 does 966, J2 does 696 (just like 966).
The pattern is 966 in which J1 and J2 do 966 (lofty double pass, self, self).
example 2 : 96677 (asynchronous bookends)
The 4handed siteswap is 9 6 6 7 7 9 6 6 7 7 ...
J1 does 96767, J2 does 67967 (identical to 96767).
The pattern is 96677 in which J1 and J2 do 96767 (lofty double pass, self, lofty single pass, self, lofty single pass).
example 3 : 9629669669969929 (Copenhagen countdown)
The 4handed siteswap is 9 6 2 9 6 6 9 6 6 9 9 6
9 9 2 9
... J1 does 92696992, J2 does 69669699.
The pattern is 9629669669969929 in which J1 does 92696992 and J2 do 69669699. Don't feel obligated to try it, it's just to have an example where J1 and J2 don't do the same thing (this is because the length of the sequence is an even number).