Single node typesense instance on docker fails to restart

I have a single node Typesense instance running on docker (following the getting started docs) that I inserted a modest number of documents to. After a restart though, this instance fails to start. It takes a minute or so to load all the previously indexed docs and then abruptly quits.

docker run -p 8108:8108 -v$(pwd)/typesense-data:/data typesense/typesense:28.0 \
  --data-dir /data --api-key=$TYPESENSE_API_KEY --enable-cors

Relevant part of the logs

I20250316 08:10:06.572754   556 collection_manager.cpp:425] Loaded 3 collection(s) so far
I20250316 08:10:06.619529   534 collection_manager.cpp:509] Loaded 3 collection(s).
I20250316 08:10:06.624069   534 raft_server.cpp:596] Finished loading collections from disk.
I20250316 08:10:06.635862   534 raft_server.cpp:607] Loaded 0conversation model(s).
I20250316 08:10:06.636823   534 raft_server.cpp:611] Initializing batched indexer from snapshot state...
I20250316 08:10:06.643399   534 batched_indexer.cpp:595] req_res.start_ts: 1742110003399726, req_res.next_chunk_index: 1
I20250316 08:10:06.646940   534 batched_indexer.cpp:624] Restored 1 in-flight requests from snapshot.
I20250316 08:10:06.658744   534 raft_server.cpp:624] Loaded 0 personalization model(s).
I20250316 08:10:06.664312   534 raft_server.h:294] Configuration of this group is 172.17.0.2:8107:8108
I20250316 08:10:06.699020   534 snapshot_executor.cpp:264] node default_group:172.17.0.2:8107:8108 snapshot_load_done, last_included_index: 4533 last_included_term: 2 peers: "172.17.0.2:8107:8108"
I20250316 08:10:06.750931   406 raft_meta.cpp:521] Loaded single stable meta, path /data/state/meta term 4 votedfor 172.17.0.2:8107:8108 time: 22506
I20250316 08:10:06.787732   406 node.cpp:608] node default_group:172.17.0.2:8107:8108 init, term: 4 last_log_id: (index=4546,term=4) conf: 172.17.0.2:8107:8108 old_conf:
I20250316 08:10:06.836714   406 node.cpp:1645] node default_group:172.17.0.2:8107:8108 term 4 start vote and grant vote self
I20250316 08:10:06.987057   406 raft_meta.cpp:546] Saved single stable meta, path /data/state/meta term 5 votedfor 172.17.0.2:8107:8108 time: 78848
I20250316 08:10:07.079692   406 node.cpp:1899] node default_group:172.17.0.2:8107:8108 term 5 become leader of group 172.17.0.2:8107:8108
I20250316 08:10:07.137110   406 raft_server.cpp:136] Node last_index: 4546
I20250316 08:10:07.140154   406 typesense_server_utils.cpp:298] Typesense peering service is running on 172.17.0.2:8107
I20250316 08:10:07.140748   406 typesense_server_utils.cpp:299] Snapshot interval configured as: 3600s
I20250316 08:10:07.141906   406 typesense_server_utils.cpp:300] Snapshot max byte count configured as: 4194304
W20250316 08:10:07.142136   406 controller.cpp:1550] SIGINT was installed with 1
I20250316 08:10:07.154639   406 raft_server.cpp:683] Term: 5, pending_queue: 0, last_index: 4547, committed: 4547, known_applied: 4533, applying: 0, pending_writes: 0, queued_writes: 1, local_sequence: 2089607
W20250316 08:10:07.159222   406 node.cpp:843] [default_group:172.17.0.2:8107:8108 ] Refusing concurrent configuration changing
E20250316 08:10:07.165160   539 raft_server.h:62] Peer refresh failed, error: Doing another configuration change
I20250316 08:10:09.855855   534 raft_server.h:294] Configuration of this group is 172.17.0.2:8107:8108
I20250316 08:10:09.908322   534 raft_server.h:294] Configuration of this group is 172.17.0.2:8107:8108
I20250316 08:10:09.934266   534 raft_server.h:294] Configuration of this group is 172.17.0.2:8107:8108
I20250316 08:10:09.955137   534 node.cpp:3298] node default_group:172.17.0.2:8107:8108 reset ConfigurationCtx, new_peers: 172.17.0.2:8107:8108, old_peers: 172.17.0.2:8107:8108
I20250316 08:10:09.982980   534 raft_server.h:277] Node becomes leader, term: 5

Reproduce the restart with the provided Docker command and persisted /data volume, then inspect the startup path in collection_manager.cpp, raft_server.cpp, batched_indexer.cpp, snapshot_executor.cpp, raft_meta.cpp, and node.cpp. Use the logged configuration-change and peer-refresh messages as the starting point. Done means a single-node container restarts successfully and serves its previously indexed documents without exiting.