So, i've been reading volume 1 of Polchinski's String Theory text book and have a doubt. His first derivation of the Weyl anomaly goes as follows:
From dimensional analysis, we know that:
$$\begin{align}T_{z\bar{z}} = \frac{a_1}{2}g_{z\bar{z}}R.\end{align}\tag{3.4.9}$$
Where $T_{ab}$ is the energy-momentum tensor and $R$ is the scalar curvature of the worldsheet. Taking covariant derivatives and using the conservation of $T_{ab}$, it is direct to show that $$ \nabla^zT_{zz}=-\frac{a_1}{2}\partial_zR. \tag{3.4.11}$$
We can now fix the constant $a_1$ by comparing the Weyl transformations in both sides. Expanding around a flat world-sheet, the RHS becomes $$ a_1\partial_z\nabla^2\delta\omega \approx 4a_1\partial^2_z\partial_\bar{z}\delta\omega. \tag{3.4.12}$$
Moreover, from the $TT$ OPE one shows that, under conformal transformations, $T_{zz}$ transforms as
$$ \epsilon^{-1}\delta T_{zz} = -\frac{c}{12}\partial^3_z v^{z}(z) - 2\partial_z v^z(z)T_{zz}(z)-v^z(z)\partial_zT_{zz}(z). \tag{3.4.13}$$
Now, from previous considerations in the chapter (section 3.3) we know that this conformal transformation is equivalent to a coordinate transformation $\delta z = \epsilon v(z)$ plus a Weyl transformation $$2\delta\omega = \epsilon\partial_z v(z)+(\epsilon\partial_z v(z))^*.$$ Note that the last two terms in the equation above are just the tensor transformation under diffeomorphisms, which means that, under Weyl transformations, $T_{zz}$ transforms as $$ \delta_W T_{zz} = -\frac{c}{12}\partial^3_zv(z) = -\frac{c}{6}\partial_z^2\delta\omega,\tag{3.4.14}$$
where the fact that $v$ is holomorphic was used in the second equality. Now, since we are expanding around a flat metric, equating the Weyl tranformation of both sides of equation (3.4.11) gives
$$\begin{align} &-\frac{c}{6}\partial^z\partial^2_z\delta\omega = 4a_1\partial^2_z\partial_{\bar{z}}\delta\omega \\ &-\frac{c}{3}\partial_\bar{z}\partial^2_z\delta\omega = 4a_1\partial^2_z\partial_{\bar{z}}\delta\omega, \end{align}\tag{2}$$
and this would imply that $$ a_1 = -\frac{c}{12}, $$ which concludes his derivation.
But, in the derivation above, we used a $\delta\omega$ which is the sum of a holomorphic function with an anti-holomorphic one (remember that $2\delta\omega = \epsilon\partial_z v(z)+(\epsilon\partial_z v(z))^*$). Doesn't this means that $$ \partial_z\partial_\bar{z}\delta\omega = 0 $$ and, thus, equation (2) is satisfied regardless of the relation between $a_1$ and $c$ (since the LHS and RHS are zero)? It seems to me that this fact spoils the demonstration. Does it? If not, why?