There is a complete characterization of equilibria in Malueg (2010) [1].
The structure of these equilibria is in a sense a generalization of the property mentioned in the question - that demands should sum up to 100. Instead for every demand $x_s$ that Sarah makes with positive probability, there is some probability that Ruth would demand exactly the remaining $100 - x_s$.
Formally, suppose one player mixes over some set A, and the other - over set B. There is a unique equilibrium for any nonempty closed A and B such that $1 \notin A\cup B$ iff these sets are balanced, i.e:
$$ s \in A \iff (100-s) \in B$$
Sets A and B do not need to be identical or finite. Moreover, these are the only equilibria (see Proposition 1 in the paper).
The equilibria indeed contain an atom. For a concrete example (example 3 from the paper) take any $a \in (0,1/2]$ and $A = B = [a,1-a]$. Each player completely mixes over this set with CDF:
$$F(s)=\begin{cases}
0 & \text{if } s<a \\
\frac{a}{1-s} & \text{if } a \le s \le 1-a \\
1 & \text{if } 1-a<s \le 1
\end{cases}$$
That is, with probability $a/(1-a)$ the player bids $x_i=a$, the lowest demand in $A$, and otherwise demands more, say $s$. If she does demand more there is a risk of running over a 100 if the other player demands more than $1-s$, which occurs with probability exactly $1-F(1-s)$ by symmetry. Then we can check that all actions in the support do bring equal payoffs as necessary for the Nash equilibrium, namely $s \times F(1-s) = a$.
[1]: Malueg, David A. "Mixed-strategy equilibria in the Nash Demand Game." Economic Theory 44.2 (2010): 243-270. It is in open access - https://core.ac.uk/download/pdf/81838757.pdf