Network may be partioned, skip forgeting ingesters this round

Hello.
My question is related to issue 3360

I have two hosts with Loki in monolithic mode. KV store is Consul.

ingester:
  lifecycler:
    ring:
      kvstore:
        store: consul
        consul:
          host: localhost:8500
      heartbeat_timeout: 20s
      replication_factor: 1
  autoforget_unhealthy: true

I turn off one host and expect the remaining Loki to forget the unhealthy instance. But it doesn’t. In the logs I see the message:
autoforget have seen 1 unhealthy ingesters out of 2, network may be partioned, skip forgeting ingesters this round
The instance remains on the list with the “unhealthy” status. Receiving logs does not work.

Please explain to me why this is happening and how I can make the remaining Loki work.
Thanks!