patroni: A cluster with a single "nofailover" node left will never recover

Describe the bug If node A tagged nofailover is the only node left in a cluster, and then you bring back node B, node B will never become a leader until you failover the cluster. But even you failover it manually, it remains broken.

To Reproduce

1. Create a cluster of 4 nodes: vas2, vas3, vas4, vas5
2. Tag 4 with nofailover
3. Poweroff nodes 2, 3, 5
4. Watch 4 remaining a single Replica with no Leader in cluster (expected)
5. Restore power to 3 and see that it never becomes Leader (unexpected)
6. Run failover manually to recover and watch that 3 has become Leader but timelines have diverged:

This is probably a broken cluster.

Expected behavior I expect 3 to become the new Leader immediately after poweron and 4 to start following it.

Environment

• Patroni version: 2.1.2-2.pgdg100+1
• PostgreSQL version: 13.6
• DCS (and its version): Consul v1.11.3
• OS: Debian GNU/Linux 10 (buster)

patronictl show-config

loop_wait: 10
master_start_timeout: 300
maximum_lag_on_failover: 1048576
postgresql:
  parameters:
    max_connections: 960
    max_locks_per_transaction: 256
    max_prepared_transactions: 960
    max_replication_slots: 10
    max_wal_senders: 10
    max_worker_processes: 16
    track_commit_timestamp: 'on'
    wal_keep_size: 16GB
  pg_hba:
  - local   all             all                               trust
  - host    all             all       0.0.0.0/0               md5
  - local   replication   repmgr                              trust
  - host    replication   repmgr      0.0.0.0/0               md5
  remove_data_directory_on_diverged_timelines: false
  remove_data_directory_on_rewind_failure: true
  use_pg_rewind: false
retry_timeout: 10
synchronous_mode: true
synchronous_node_count: 3
ttl: 30

Have you checked Patroni logs?

The unfortunate nofailover node writes:

INFO: no action. I am (vas4), a secondary, and following a leader (vas5)

(vas5 is not even there, it was powered off).