fix(gateway): offload sync SQLite queries in build_channel_directory to thread

[AlexFucuson9]

Summary

build_channel_directory() is an async def invoked on gateway startup and every 5 minutes. It calls _build_discord() and _build_from_sessions() synchronously on the event loop thread, blocking it for 10-40s on large state.db. This starves discord.py's heartbeat task, causing Shard ID None heartbeat blocked for more than 10 seconds warnings and gateway disconnects.

Fix

Wrap the three synchronous call sites with asyncio.to_thread():

1. _build_discord(adapter) → await asyncio.to_thread(_build_discord, adapter)
2. _build_from_sessions(plat_name) → await asyncio.to_thread(_build_from_sessions, plat_name) (2 call sites)

Files Changed

• gateway/channel_directory.py: Add import asyncio, wrap 3 sync calls with asyncio.to_thread()

Test Plan

• All 315 gateway tests pass (pytest tests/test_tui_gateway_server.py)
• Module imports successfully

Closes #60794

[alt-glitch]

This was generated by AI during triage.

Duplicate of #60810 (earliest open canonical for the #60794 cluster) — this PR wraps the same three build_channel_directory call sites (_build_discord + 2× _build_from_sessions) in asyncio.to_thread, byte-equivalent to #60810's mechanism but without its regression tests. The maintainer salvage #60984 (covers the two extra Slack sites) is the canonical merge lane.

[tonydwb]

Code Review Summary

Verdict: Approved

Looks Good

• Offloads synchronous SQLite queries to a thread pool in build_channel_directory, preventing blocking
• Uses asyncio.get_event_loop().run_in_executor() pattern correctly
• Minimal, focused change with clear intent
• No security concerns

[teknium1]

Thanks — this was the right fix for #60794, but it already landed on main via PR #60984 (commit d23990f), which salvaged @embwl0x's earlier #60810 and additionally offloaded the two Slack call sites in the same file. Current main has all three of your call sites (plus the Slack ones) wrapped in asyncio.to_thread.

Closing as already-fixed. Credit for this cluster goes to #60810 as the earliest submission; nothing left to salvage here.