Bug: nexusd can fail on startup due to attempting to use default pkarr relays when in testnet mode

[catch-21]

pubky-nexus version: b0d428b
pubky-stack version: https://github.com/pubky/pubky-stack/tree/ce44db8342d7e0673eba8f6862ba11e163a1ff9b

Using nexusd in our E2E test pipeline works most of the time. One occurrence of it failing startup has been noticed here https://github.com/pubky/franky/actions/runs/18143567883/job/51640128555 (look at 'Dump docker logs on failure' step).

2025-09-30T21:12:33.175256Z DEBUG hyper_util::client::legacy::pool:395: pooling idle connection for ("https", relay.pkarr.org)
2025-09-30T21:12:33.175497Z DEBUG pkarr::client::relays:337: PUT reqwest::Error { kind: Request, url: "https://relay.pkarr.org/kspydab4rzhrxp9ydrmegp86ju3xo3n9kumeojg5rz5haaxzg65o", source: TimedOut }
2025-09-30T21:12:33.175529Z DEBUG hyper_util::client::legacy::pool:395: pooling idle connection for ("https", pkarr.pubky.org)
2025-09-30T21:12:33.175560Z DEBUG pkarr::client::relays:337: PUT reqwest::Error { kind: Request, url: "https://pkarr.pubky.org/kspydab4rzhrxp9ydrmegp86ju3xo3n9kumeojg5rz5haaxzg65o", source: TimedOut }
2025-09-30T21:12:33.175586Z  WARN nexus_webapi::key_republisher:49: Failed to publish the service's pkarr packet to the DHT: Publish query timed out with no responses neither success or errors.
2025-09-30T21:12:33.175643Z ERROR nexus_webapi::builder:86: Failed to start Nexus API: Publish query timed out with no responses neither success or errors.

Config being used is https://github.com/pubky/pubky-docker/blob/70054e36f8aa63e9f63c2ff579614e5ef33fc8ce/pubky-nexus-config-testnet.toml

[catch-21]

Three things to notice:

1. Nexus fails to publish over the mainnet relays, it shouldn't fail, but it does. There should be internet in this environment, so I'm unsure why this failed.
2. Nexus is trying to publish to mainnet, but should be doing that on Testnet.
3. Nexus should still start in this scenario. Log error and proceed.

[catch-21]

again https://github.com/pubky/franky/actions/runs/18849758754/job/53783142848#step:22:465

[ok300]

I am also occasionally seeing this in bare-bones simple testnet setups (loop that signs up users on a local testnet environment, with no Nexus involved).

[ok300]

This is thrown by the mainline library here:

https://github.com/pubky/mainline/blob/4062830a4ee1240eec45bcc88433972c75fa9d57/src/rpc/put_query.rs#L138

[ok300]

I think this was fixed by pubky/mainline#78

Testnet startup latency is also due to Testnet::new blocking on bootstrapped() for each node, forcing network lookups and timeout-driven waits. This is however unnecessary for local testnets where all nodes are known in-process. The change avoids those waits while keeping production DHT behavior unchanged.
pubky/mainline#78 (comment)

[catch-21]

@ok300 Can you let me know when that is in nexus so I can use it for E2E tests.

[ok300]

@catch-21 it looks like this fix will come with #726 , which includes Pubky SDK v0.7 and the latest mainline, pkarr.