$LimitTo4 is a variable with default setting False. If set to True, the limit Dimension -> 4 is performed after tensor integral decomposition.
$LimitTo4 is a global variable that determines whether UV-divergent Passarino-Veltman functions are simplified by taking the limit D-4 \to 0.
A generic IR-finite Passarino-Veltman function X can be written as X = \frac{a}{D-4} + b + \mathcal{O}(\varepsilon), with a being the prefactor of the pole and b being the finite part. Therefore, products of such functions with coefficients that are rational functions of D with f(D) = f(4) + (D-4) f'(4) + \mathcal{O}(\varepsilon^2) can be simplified to f(D) X = f(4) X + a f'(4) + \mathcal{O}(\varepsilon), whenever such products appear in the reduction.
This relation is correct only if the Passarino-Veltman functions have no IR divergences, or if such divergences are regulated without using dimensional regularization.
For this reason, even when $LimitTo4 is set to True, the simplifications are applied only to A and B functions. Although B functions can exhibit an IR divergence, such integrals are zero in dimensional regularization, so that no mixing of \varepsilon-terms from IR and UV can occur.
The default value of $LimitTo4 is False. Notice that even when the switch is set to True, it will essentially affect only the Passarino-Veltman reduction via PaVeReduce.
The modern and more flexible way to simplify amplitudes involving IR-finite PaVe functions is to use the special routine PaVeLimitTo4.
Overview, PaVe, PaVeReduce, OneLoop, $LimitTo4IRUnsafe, PaVeLimitTo4.
$LimitTo4\text{False}