Our solution is to extend timeouts if the situation looks dire. More
precisely, if a node does not have a finalized block in Δstandstill ≈ 10 sec of
consecutive leader windows, the node will start extending its timeouts by
ε ≈ 5% in every leader window.
The skip timer should be made aware of when we enter / exit standstill so that it can adjust timeouts accordingly
From the paper:
The skip timer should be made aware of when we enter / exit standstill so that it can adjust timeouts accordingly