Let's consider an example case. Take one sixth of a mole of gas particles; that's $10^{23}$ particles. The probability that all happen to be in one half of a box (assuming equal probabilities for the two halves for each particle) is
$$
2^{-10^{23}}.
$$
Suppose each particle takes a millisecond to cross the box, so that the whole gas takes about 1 millisecond to sample a new distribution across the two halves of the box. The number of distributions thus sampled per year is
1 year / 1 millisecond $\simeq 3 \times 10^{10}$
So the expected number of years until a gathering all in one half of the box occurs is
$$
T = 2^{10^{23}} / 3\times 10^{10}.
$$
To understand this number, let's take the log to base 10:
$$
\log_{10} T = 10^{23} \log_{10} 2 - \log_{10} 3 - 10 \simeq 0.3 \times 10^{23}
$$
Now compare with the age of the universe, which I will take as 14 billion years:
$$
\log_{10} (T / 1.4 \times 10^{10} ) \simeq 0.3 \times 10^{23} - \log_{10} 1.4 - 10 \simeq 0.3 \times 10^{23}.
$$
Now to be cautious let's take a third of this, thus getting $10^{22}$. So the estimate is that it will take a time greater than
$10^{10^{22}}$ times the current age of the universe from the Big Bang
for the gas to gather spontaneously in one half of the box, if it is to happen just by random independent motion of each of the particles.
None of our knowledge of physics can be trusted on this large a timescale. What the calculation really means is that the authors were incorrect to assert that the gas "could" spontaneously gather in one side of the box. What the calculation means is that the gas could not gather on one side, merely by independent random motions of the particles, by any reasonable definition of the words "could" and "could not".