In [1] the following edge-connectivity related problem is studied.

Consider the following LP relaxation,

$$\begin{array}{rlrl}\min & \sum _{e\in E}w(e)x_e& & \\ s.t.& \sum _{e\in \delta (S)}{x}_e\ge k& & \forall S\subset V\\ & 0\le x_e\le 1& & \forall e\in E\end{array}$$

Note that $x_e\le 1$ is necessary since each edge can only be chosen once. In the paper the authors mentioned that this kind of constraints are called box constraints and they usually make LPs difficult to solve (for example, positive covering LPs can be solved through MWU). There is also a box-free version of LP relaxation,

$$\begin{array}{rlrl}\min & \sum _{e\in E}w(e)x_e& & \\ s.t.& \sum _{e\in C\setminus S}{x}_e\ge k-|S|& & \forall \text{ cut }C\\ & & & \forall S\in \left\{F:|F|\le k-1\wedge F\subset C\right\}\\ & x_e\ge 0& & \forall e\in E\end{array}$$ which is box-free but includes exponentially many extra constraints. I will call the first LP boxLP and the second one boxlessLP. Any feasible solution to the boxLP is also feasible in the boxlessLP. For any $x_e>1$ in a feasible solution of boxlessLP, we consider those cuts containing $e$. For such a cut $C$, ${\sum }_{f\in C\setminus e}{x}_f\ge k-1$ and thus ${\sum }_{e\in C}{x}_e>1$. Since this holds for any cut $C$ containing $e$, we can certainly decrease $x_e$ for a smaller objective. Hence, in the optimal solution to boxlessLP, every $x_e$ is less than or equal to 1.

In fact, one can see from the proof that enumerating all singletons $F=\left\{f\right\}$ is sufficient.

References