I'm currently reading these notes about the Ward identity (pages 259 - 261). I will repeat some of the steps to make the question self-contained.
Let us consider a local transformation on the field $\phi$: \begin{equation} \phi(x) \rightarrow \phi'(x) = \phi(x) + \delta \phi(x) \tag{1} \end{equation} where: \begin{equation} \delta \phi(x) = i \epsilon^a(x) t^a \phi(x) \end{equation} where $t^a$ are the generators of the transformation and $\epsilon$ is a the space-time dependent parameter characterizing the field transformation. Then, the Noether current is given by: \begin{equation} j_\mu^a = i \frac{\partial \mathcal{L}}{\partial \left(\partial_\mu \phi \right)} t^a \phi \end{equation} and the variation of the action is: \begin{equation} \delta S = \int \mathrm{d}^4 x \; j_\mu^a \partial^\mu \epsilon^a \end{equation} provided $\delta S = 0$ for global transformations.
Let us now consider the usual generating functional:
\begin{equation} Z[J] = \int \mathcal{D}\phi \; \exp\left(i S \left[\phi, \partial_\mu \phi \right] + i \int\limits \mathrm{d}^4 x \; J \phi \right) \end{equation}
If we subsequently perform the (local) change of variables (see equation $(1)$), and assume that the integration measure is invariant, then we get:
$$ Z[J] = \int \mathcal{D}\phi \; \exp\left(i S \left[\phi, \partial_\mu \phi \right] + i \delta S \left[\phi, \partial_\mu \phi \right] + i \int\limits \mathrm{d}^4 x \; J \phi - \int\limits \mathrm{d}^4 x \; J \epsilon^a t^a \phi \right)$$ $$ = \int \mathcal{D}\phi \; \exp\left(i S \left[\phi, \partial_\mu \phi \right] + i \int \mathrm{d}^4 x \; j_\mu^a \partial^\mu \epsilon^a + i \int\limits \mathrm{d}^4 x \; J \phi - \int\limits \mathrm{d}^4 x \; J \epsilon^a t^a \phi \right) \tag{2}$$
This is equation (10.170) in the aforementioned notes (up to a minus sign). Now, according to the same notes, we can expand the above equation at the first order in $\epsilon^a$ as follows: \begin{equation} Z[J] = \int \mathcal{D}\phi \; \exp\left(i S \left[\phi, \partial_\mu \phi \right] + i \int\limits \mathrm{d}^4 x \; J \phi \right)\left(1 + i \int \mathrm{d}^4 x \; j_\mu^a \partial^\mu \epsilon^a - \int\limits \mathrm{d}^4 x \; J \epsilon^a t^a \phi \right) \tag{3} \end{equation} Now, my question is:
To get from equation $(2)$ to $(3)$: why can we just use normal Taylor series to get expand the exponential? Shouldn't we be using the Baker–Campbell–Hausdorff formula?
Of course, if we are only considering Abelian gauge transformations, then we can just use the normal Taylor series. However, they are not mentioning this and now I am confused.
