I'm new to the concept of finding a limit using a Taylor polynomial. I am curious when is it ok to stop adding terms to the polynomial if I am trying to find a limit? I've seen examples where people stopped at the $x^2$ term and then added $O(x^3)$ to the expression, examples where people went as far as $x^7$ and then added $O(x^8)$ to the expression and everything in between. If what I want is to find a limit, how should I know when to stop? What would be acceptable in say, an exam situation?
EDIT
For example, say I am trying to find the limit
$$\lim_{x \to 0} \dfrac{\sin x}{x}$$
without using L'Hospital. So I am trying to use the Taylor Expansion. By "When is it acceptable to stop adding terms" I'm really asking about what is the difference between doing something like:
$$\lim_{x \to 0} \dfrac{x - \frac{x^3}{3!} + O(x^5)}{x}$$
or doing something like:
$$\lim_{x \to 0} \dfrac{x - \frac{x^3}{3!} + \frac{x^5}{5!}+ O(x^7)}{x}$$
or something like:
$$\lim_{x \to 0} \dfrac{x - \frac{x^3}{3!} + \frac{x^5}{5!} - \frac{x^7}{7!} + O(x^9)}{x}$$
and so on. Which one should I use? That's what I mean when I say "When is it acceptable to add terms".
There is no hard and fast rule for the number of terms you need and it will depend on context and the properties of the function in the limit. I've tried to come up with a pathological example to demonstrate how the degree can be hard to pin down.
Consider the Taylor expansion $\displaystyle \ln\left(1-x\right)=-\sum_{n=1}^{\infty}\frac{x^{n}}{n}$, which is valid for all $x\in[-1,1)$. We should have $\displaystyle L=\lim_{x\to1^-}\frac{\ln (1-x)}{\ln (1-x)}=\lim_{x\to1^-}1=1$. However, if we try to evaluate the limit with Taylor polynomials, we would have
$$\displaystyle L\stackrel{?}{=}\lim_{x\to1^-}\frac{-\sum\limits_{n=1}^{N_{1}}\frac{x^{n}}{n}}{-\sum\limits_{n=1}^{N_{2}}\frac{x^{n}}{n}}=\lim\frac{x+\frac{x^2}{2}+\ldots+\frac{x^{N_1}}{N_1}}{x+\frac{x^2}{2}+\ldots+\frac{x^{N_2}}{N_2}}$$
for respective upper bounds $N_1, N_2$. The two sums are finite and continuous at $1$ so we can substitute $x=1$. We can then use a property of the Harmonic numbers and Euler's constant, $\gamma=0.577$, to approximate the expression as
$$\displaystyle L\stackrel{?}{=}\frac{\sum\limits_{n=1}^{N_1}\frac{1}{n}}{\sum\limits_{n=1}^{N_{2}}\frac{1}{n}}\approx \frac{\gamma+\ln N_1}{\gamma+\ln N_{2}}$$
so in fact, despite knowing that $L=1$, we can make this expression as large or as small as we want by choosing the right $N_1, N_2$. We also see that we can have arbitrarily large $N_1,N_2$ that give us the wrong answer. The take-home message is that naively evaluating limits by substituting a Taylor polynomial can lead to incorrect results, even if you use one with a high degree.
In the figure below, we have $\ln (1-x)$ in dotted black, which is well approximated over $[-1,1)$ by the Taylor polynomials up to $N_1=10$ (red curve) and $N_2=5$ (blue). However, the true value of $\displaystyle \frac{\ln (1-x)}{\ln (1-x)}$ is the purple horizontal line equal to $1$. This does not have the same limit at $1$ as the ratio of Taylor polynomials (green), which is approximately $\displaystyle \frac{\gamma+\ln 10}{\gamma+\ln 5}=1.317$ (red point). You can try this for yourself with this (Desmos graph).