We know that the class equation of $A_4$, the alternating group of $S_4$, is $1+3+4+4$.
The order of $cl((1))$ is $1$, and order of $cl((12)(34))$ is $3$. And we split the rest of the $8$ elements into $cl((123))$ and $cl((132))$, and both of these have order $4$.
My question is, since $2$ and $6$ can still divide the order of the group $A_4$, why cannot these $8$ elements be divided into two conjugate classes with each orders of $2$ and $6$?
Let $H$ be a normal subgroup of $G$. If $C \subset H$ is an $H$-conjugacy class, then $gCg^{-1} \subset H$ is also an $H$-conjugacy class, so they will have the same order. To see this, it suffies to show that: $$\text{If} \ c \in C, \ \text{so} \ gcg^{-1} \in gCg^{-1}, \text{then} \ h(gcg^{-1})h^{-1} \in gCg^{-1}.$$
But we have
$$h(gcg^{-1})h^{-1} = g g^{-1} (h(gcg^{-1})h^{-1}) g g^{-1} = g (g^{-1} h g) c (g^{-1} h^{-1} g) g^{-1}= g \gamma c \gamma^{-1} g^{-1},$$ where $\gamma = g^{-1} h g \in H$. Since $\gamma c \gamma^{-1} \in C$, we are done.
Thus $G$ acts on the conjugacy classes of $H$ by conjugation. Certainly the union of $gCg^{-1}$ for a fixed $C$ over all $g \in G$ is just the $G$-conjugacy class of any $c \in C$. But clearly $C$ and $gCg^{-1}$ have the same order. Hence a conjgugacy class of $G$ inside $H$ breaks up into $H$-conjugacy classes all of the same order. Since $G$ permutes them and $H$ fixes them, the $G$ action factors through $G/H$, and thus the number of pieces has to divide $G/H$.
If $H = A_n$ and $G=S_n$ then $G/H$ has order $2$, so any $S_n$-conjugacy class inside $A_n$ breaks up into either $1$ or $2$ $A_n$-conjugacy classes of the same order.