Question about LeaseRead

In the Raft paper, there is a mentioned optimization method for Lease Read using clock + heartbeat. When the leader sends a heartbeat, it first records a timestamp called "start". If the majority of the nodes in the system respond with heartbeat responses, it is assumed that the leader's lease is valid until the time point "start + election timeout / clock drift bound".
But when I read etcd LeaseRead code, it seems that LeaseRead depend on the checkquorum. When the leader receive the major of heartbeat responses in the half of election_timeout, the MsgCheckQuorum will not step the leader down on this election_timeout round. The next round, the leader step down into follower because the major of nodes is not recentActive.This case seems that the leasetime last 1.5 election_timeout.This confused me.

Start by reading the LeaseRead and checkquorum handling, especially the MsgCheckQuorum path and heartbeat response timing mentioned in the issue. Trace how recentActive is updated across election-timeout rounds, then document whether the observed lease duration matches the intended behavior or identifies a defect.

Written by the indexing model from the issue text.