Waiting two ttl-periods before declaring a lease lost seems wrong

I was reading the code to understand the lease refresh algorithm and notice this:

    // When the cluster goes down, we keep trying to reconnect. But if we're
    // far past the end of our key's TTL, there's no way we're going to be
    // able to renew it. Fire a "lost".
    if (Date.now() - this.lastKeepAlive > 2 * 1000 * this.ttl) {
      ...
    }

Waiting 2 *ttl seem problematic, the process will believe it has the lease a full ttl period after it has certainly lost it. Can this be configurable somehow? Why isn't it just 1 period?

Start with the lease refresh algorithm and the code that compares Date.now() - this.lastKeepAlive with 2 * 1000 * this.ttl. Trace how lease loss is detected and confirm the intended TTL behavior, including whether the threshold should be configurable. Done means the lease-loss behavior is clearly defined and covered by the relevant tests, if present.