Description

• When Telegram hits a transient network failure during startup, Hermes treats it as a startup failure and can stop before the background reconnection watcher has a chance to recover.

Expected Behavior

• If the Telegram failure is marked retryable (for example, temporary DNS/connect errors), the gateway should stay running and keep retrying in the background until the network returns.

Actual Behavior

• Telegram connect errors are already marked retryable.
• The gateway startup path still returns failure when no platform is connected.
• This can prevent the reconnection watcher from doing its job.

Why this looks like a bug

• The code already has retry logic for Telegram network issues, so the current startup behavior conflicts with that design. It behaves like a temporary outage becomes a hard failure.

Reproduction

• Start Hermes with Telegram enabled. Trigger a transient Telegram network/connectivity failure. Observe that startup can fail instead of staying up and retrying.