Technically there is a way to go through this just in terms of algebra, but I think it is easier to just talk about an example. The simplest case where this comes up that I can think of is:
$$\frac{1}{x(x^2+1)}=\frac{1}{x}-\frac{x}{x^2+1}.$$
You might guess $\frac{1}{x(x^2+1)}=\frac{A}{x}+\frac{B}{x^2+1}$. One clue that this can't be right is that it has the wrong scaling: the left side behaves like $x^{-3}$ as $x \to \infty$ while the right side can only behaves as $x^{-1}$ as $x \to \infty$. There is no way the second term can "cancel out enough" of the first term to give the $x^{-3}$ scaling.
But let's say you don't know anything about scaling and just want to think about coefficients. You clear denominators and get
$$1=A(x^2+1)+Bx$$
How will the $x^2$ and $x$ terms be eliminated? There is no way, short of $A=B=0$ which is also incorrect. But if $B$ were instead $Bx+C$, then you would have a $Bx^2$ term there to be able to cancel out the $Ax^2$. Conversely, just $Bx$ would not be sufficient if the numerator were not constant, because then there would be no linear term to balance the linear term on the left side of the equation. In general you need both.
A way to see it as not being a special case is to go to complex numbers. Note that this is usually more trouble than it's worth when actually trying to find antiderivatives of rational functions because of issues with ambiguity of the complex logarithm. But for illustration it helps.
You have:
$$\frac{1}{x(x^2+1)}=\frac{A}{x}+\frac{B}{x-i}+\frac{C}{x+i}$$
for complex constants $A,B,C$. This is the same as the usual non-repeated linear factor case. You could compute $A,B,C$ directly and then simplify, but you can also get a common denominator in the second and third terms while still leaving everything symbolic:
$$\frac{B}{x-i}+\frac{C}{x+i}=\frac{B(x+i)+C(x-i)}{x^2+1}=\frac{(B+C)x+i(B-C)}{x^2+1}.$$
Thus $\frac{B}{x-i}+\frac{C}{x+i}=\frac{Dx+E}{x^2+1}$ where $D=B+C,E=i(B-C)$. Note that a linear term appears when you try to force two terms with real roots to have a common denominator, for exactly the same algebraic reason. For example:
$$\frac{1}{x-1}+\frac{1}{x+2}=\frac{x+2+x-1}{(x-1)(x+2)}=\frac{2x+1}{(x-1)(x+2)}.$$
From this perspective the linear term shows up because in choosing to do the decomposition over $\mathbb{R}$ we are forced to combine two terms in the decomposition over $\mathbb{C}$ that have different roots.