Now $\xi^{G\setminus Y}$ is a valid path of the Markov Chain associated to the graph $G\setminus Y$ (i.e. with vertices $X\cup S$), because in the original path $\xi^G$, all zeroes in $X$ have been resampled by resamplings in $X$. There can not be a vertex $v\in Y$ such that the resampling of $v$ changed a vertex in $X$, since $X$ and $Y$ are only connected through $S$ and we know $\xi^G\in\NZ{S}$.

    Vice versa, any two paths $\xi^{G\setminus Y}\in\NZ{S}$ and $\xi^{G\setminus X}\in\NZ{S}$ also induce a path $\xi^G\in\NZ{S}$ by simply interleaving the resampling positions. Note that $\xi^{G\setminus Y},\xi^{G\setminus X}$ actually induce $\binom{|\xi^{G\setminus Y}|+|\xi^{G\setminus X}|}{|\xi^{G\setminus Y}|}$ paths $\xi^G$ because of the possible orderings of interleaving the resamplings in $\xi^{G\setminus Y}$ and $\xi^{G\setminus X}$.

    For a fixed $\xi^{G\setminus Y},\xi^{G\setminus X}$ we will now show the following:

    \begin{align*}

        \sum_{\substack{\xi^G\in\NZ{S} \text{ s.t.}\\ \xi^G \text{ decomposes into } \xi^{G\setminus Y},\xi^{G\setminus X} }} \P^{G}_S(\xi^G) &=

        \sum_{\text{interleavings of }\xi^{G\setminus Y},\xi^{G\setminus X}} \P(\text{interleaving}) \cdot \P^{G\setminus Y}_S(\xi^{G\setminus Y}) \cdot \P^{G\setminus X}_S(\xi^{G\setminus X}) \\

        &= \P^{G\setminus Y}_S(\xi^{G\setminus Y}) \cdot \P^{G\setminus X}_S(\xi^{G\setminus X}) 

    \end{align*}

    where both sums are over $\binom{|\xi^{G\setminus Y}|+|\xi^{G\setminus X}|}{|\xi^{G\setminus Y}|}$ terms.

    This is best explained by an example. Lets consider the following fixed $\xi^{G\setminus Y},\xi^{G\setminus X}$ and an example interleaving where we choose vertices from $Y,X,X,Y,\cdots$:

    \begin{align*}

        \xi^{G\setminus Y} &= \left( (\text{initialize to }b^X\;1^S),

        (z^X_1, v^X_1, r^X_1),

        (z^X_2, v^X_2, r^X_2),

        (z^X_3, v^X_3, r^X_3),

        (z^X_4, v^X_4, r^X_4),

        \cdots  \right) \\

        \xi^{G\setminus X} &= \left( (\text{initialize to }1^S\;b^Y),

        (z^Y_1, v^Y_1, r^Y_1),

        (z^Y_2, v^Y_2, r^Y_2),

        (z^Y_3, v^Y_3, r^Y_3),

        (z^Y_4, v^Y_4, r^Y_4),

        \cdots  \right) \\

        \xi^G             &= \big( (\text{initialize to }b^X \; 1^S \; b^Y),

        (z^X_1+z^Y_1, v^Y_1, r^Y_1),

        (z^X_1+z^Y_2, v^X_1, r^X_1), \\

        &\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad

        (z^X_2+z^Y_2, v^X_2, r^X_2),

        (z^X_3+z^Y_2, v^Y_2, r^Y_2),

        \cdots \big)

    \end{align*}

    Here $b^X\in \{0,1\}^{X}$ and $b^Y\in\{0,1\}^Y$. Since we condition on the event that $S$ is initialized to ones, we know the initial state is of the form $b^X\;1^S$ in $\xi^{G\setminus Y}$. Similarly, since these paths satisfy the $\NZ{S}$ event, we know all the vertices $v_i$ resampled in $\xi^{G\setminus Y}$ are vertices in $X$, and the resampled bits $r_i$ are bits corresponding to vertices in $X$.

    In the newly constructed path $\xi^G$ the number of zeroes is the number of zeroes in $X$ and $Y$ together, so this starts as $z^X_1 + z^Y_1$. Then in this example, after the first step the number of zeroes is $z^X_1+z^Y_2$ since a step of $\xi^{G\setminus X}$ was done (so a vertex in $Y$ was resampled).

    The probability of $\xi^{G\setminus Y}$ is given by

    \begin{align*}

        \P^{G\setminus Y}_S(\xi^{G\setminus Y}) &=

        \P(\text{initialize }b^X\;1^S \mid \text{initialize $S$ to }1)

        \P(\text{pick }v^X_1 \mid z^X_1) \P(r^X_1)

        \P(\text{pick }v^X_2 \mid z^X_2) \P(r^X_2) \cdots \\ 

        &= (1-p)^{|b^X|} p^{|X|-|b^X|} \cdot

        \frac{1}{z^X_1} \P(r^X_1) \cdot

        \frac{1}{z^X_2} \P(r^X_2) \cdots

        \frac{1}{z^X_{|\xi^{G\setminus Y}|}} \P(r^X_{|\xi^{G\setminus Y}|}) .

    \end{align*}

    and similar for $\xi^{G\setminus X}$.

    Instead of choosing a step in $Y,X,X,Y,\cdots$ we could have chosen other orderings. The following diagram illustrates all possible interleavings, and the red line corresponds to the particular interleaving $Y,X,X,Y$ in the example above.

    \begin{center}

    \end{center}

    For the labels shown within the grid, define $p_{ij} = \frac{z^X_i}{z^X_i + z^Y_j}$.

    The probability of this particular interleaving $\xi^G$ is given by

    \begin{align*}

        \P^{G}_S(\xi^{G})

        &= (1-p)^{|b^X\; b^Y|} p^{|X\cup Y|-|b^X\;b^Y|} \quad

        \frac{1}{z^X_1+z^Y_1} \P(r^Y_1) \cdot

        \frac{1}{z^X_1+z^Y_2} \P(r^X_1) \cdots \\

        &= (1-p)^{|b^X|} p^{|X|-|b^X|} \cdot (1-p)^{|b^Y|} p^{|Y|-|b^Y|} \\

        &\qquad \cdot

        \frac{z^Y_1}{z^X_1+z^Y_1} \frac{1}{z^Y_1} \P(r^Y_1) \;

        \frac{z^X_1}{z^X_1+z^Y_2} \frac{1}{z^X_1} \P(r^X_1) \;

        \frac{z^X_2}{z^X_2+z^Y_2} \frac{1}{z^X_2} \P(r^X_2)

        \cdots \tag{rewrite fractions}\\

        &=

        \frac{z^Y_1}{z^X_1+z^Y_1} 

        \frac{z^X_1}{z^X_1+z^Y_2} 

        \frac{z^X_2}{z^X_2+z^Y_2} 

        \cdots

        \P^{G\setminus Y}_S(\xi^{G\setminus Y}) \; \P^{G\setminus X}_S(\xi^{G\setminus X})

        \tag{definition} \\

        &= (1-p_{1,1}) \; p_{1,2} \; p_{2,2} \; (1-p_{3,2}) \cdots \P^{G\setminus Y}_S(\xi^{G\setminus Y}) \; \P^{G\setminus X}_S(\xi^{G\setminus X})

        \tag{definition of $p_{i,j}$} \\

        &= \P(\text{path in grid}) \; \P^{G\setminus Y}_S(\xi^{G\setminus Y}) \; \P^{G\setminus X}_S(\xi^{G\setminus X})

    \end{align*}

    In the grid we see that at every point the probabilities sum to 1, and we always reach the end, so we know the sum of all paths in the grid is 1. This proves the required equality.

    We obtain

    \begin{align*}

        \P^{G}_S(\NZ{S} \cap A^X \cap A^Y)

        &= \sum_{\xi^G \in \NZ{S}\cap A^X \cap A^Y} \P^{G}_S(\xi^G) \\

        &= \sum_{\xi^{G\setminus Y} \in \NZ{S}\cap A^X}

           \sum_{\xi^{G\setminus X} \in \NZ{S}\cap A^Y}

            \P^{G\setminus Y}_S(\xi^{G\setminus Y}) \cdot

            \P^{G\setminus X}_S(\xi^{G\setminus X}) \\

        &= \P^{G\setminus Y}_S(\NZ{S} \cap A^X) \; \cdot \; \P^{G\setminus X}_S(\NZ{S} \cap A^Y)

    \end{align*}

\end{proof}

\todo{rewrite from here}

\begin{lemma}[Conditional independence 2] \label{lemma:eventindependenceNewGen}

	Let $v,w \in [n]$, and let $A$ be any event that depends only on the sites $[v,w]$ (meaning the initialization and resamples) and similarly $B$ an event that depends only on the sites $[w,v]$. (For example $\mathrm{Z}^{(s)}$ or ``$s$ has been resampled at least $k$ times'' for an $s$ on the correct interval). Then we have

	\begin{align*}

		\P^{(n)}(\mathrm{NZ}^{(v,w)}\cap A\cap B)

		=

		\P_{b_v=b_w=1}^{[v,w]}(\mathrm{NZ}^{(v,w)}\cap A)

		\; \cdot \;

		\P^{[w,v]}(\mathrm{NZ}^{(v,w)}\cap B),

	\end{align*}