**General Background**

Take $G$ to be a finite group, and say $V$ is s $G$-representation over $\mathbb{Q}_p$. By picking a $G$-invariant lattice $L\subset V$ we can get an $\mathbb{F}_p$ representation by looking at $L/pL$, which we call a reduction of $V$.

It is not hard to show that if we semisimplify the reduction, we get an answer which depends only on $V$, and not on the choice of $L$. However, I am interested in how decomposable we can make the reduction. That is, I want to get as large a subrepresentation of $L/pL$ as possible which is semisimple.

**Specific Question**

Say $V$ is absolutely irreducible, and its reduction for a $G$-invariant lattice $L\subset V$ is a *non-trivial* extension of 2 simple modules $A,B$, so that we have
$$0\rightarrow A\rightarrow L/pL\rightarrow B\rightarrow 0.$$

I know that it is not always possible to pick a different $G$-invariant $L_0$ such that $L_0/pL_0=A\oplus B$.

My question is this: Can we do better by taking powers of $V$? That is, does there exist an integer $n$ and a $G$-invariant lattice $L_1\subset V^n$ such that $L_1/pL_1$ has a semisimple subrepresentation which is a summand of at least $n+1$ simple objects?

I would also appreciate any reference for studying which reductions one can get from a $\mathbb{Q}_p$ representation.

Thanks for your help!