Non-validator node stuck in bootstrap loop — P2P state transfer always times out (829MB, 60s timeout)

[Oldman-beginer]

Problem

My non-validator node crashed with "too many blocks to request" panic and lost visor_abci_state.json. Now it's stuck in
an infinite bootstrap loop — every peer times out before completing the 829MB state transfer.

Environment

• Ubuntu 24.04 LTS, 64GB RAM, NVMe SSD
• EU-based server (94.x.x.x)
• Internet: >100 Mbps (not the bottleneck)
• hl-visor: 4a5921208e0c8d246d26aeaa3958557ee538d48b (2026-02-28)
• hl-node: 105cb1dc0129b99e8a81d16b8cc4118749a0ed37 (2026-03-21)
• try_new_peers: true, 63 peers configured

What happens

1. Node connects to a peer and starts downloading abci_state (829MB)
2. Transfer speed is ~1-1.3 MB/s from all peers
3. After ~60 seconds → tcp_read_exact: deadline has elapsed (downloaded only ~60MB of 829MB)
4. Next attempt → same peer says Rate limited by peer
5. Moves to next peer → same timeout
6. Loop continues indefinitely across 100+ peers

Logs

WARN >>> hl-node @@ considering local abci state as stale @@ [stale_reason: "local height is too far behind"]
WARN >>> hl-node @@ connecting to peer: Ip(x.x.x.x)                                                                          
WARN >>> hl-node @@ reading bytes for abci_stream recv greeting: 52000000/829967179
WARN >>> hl-node @@ could not read abci state from x.x.x.x: lu::timeout tcp_read_exact: deadline has elapsed                 
WARN >>> hl-node @@ Rate limited by peer                                                                                     

What I've tried

• 63 different peers (EU, US, JP, SG) — all give ~1 MB/s, all timeout
• Helsinki Hetzner peers (65.109.x.x) — always early eof, never serve state
• Restoring from periodic_abci_states (height 939600000) — node says "too far behind", forces full bootstrap
• override_gossip_config.json with try_new_peers: true — discovers 100+ peers, none fast enough
• Running for 10+ hours — no successful bootstrap

The math

829MB ÷ 60s timeout = ~14 MB/s required. All peers serve at ~1 MB/s. This means bootstrap is physically impossible
from my location unless:

• A peer serves faster than 14 MB/s, or
• The timeout is increased, or
• State can be downloaded via HTTP/S3 instead of P2P

Questions

1. Is there a way to download abci_state.rmp directly (HTTP/S3 snapshot)?
2. Can the 60-second stream timeout be configured?
3. Is there a plan to provide state snapshots for bootstrap (like other L1 chains do)?

Related issues

• [hl-node][Mainnet] Panic: too many blocks to request in gossip forward_client_blocks after starting bootstrap; visor restarts child #141 — same "too many blocks to request" panic
• Could not read abci state from peer ips and raise early eof error #73 — same "early eof" during bootstrap (resolved by moving to Tokyo)