4 parallel research agents output (2026-05-07):
- RESEARCH-2B2T-QUEUE.md — 2b2t queue tech deep-dive: architecture, drama
timeline, 5 patterns to copy + 5 to avoid
- RESEARCH-LIMBO-PLUGIN-SURVEY.md — open-source plugin survey: STEAL list
(Elytrium LimboAPI/LimboAuth + PistonQueue), PATTERN list, SKIP list
- V2-ARCHITECTURE.md — Paper-only stack with Velocity-ready seam, 7-state
login flow, snapshot-on-pre-login, transparent FIFO trust tiers
- V2-ROADMAP.md — M0-M5 milestones with acceptance criteria + dep graph
Stack decision: Paper-only for now (no proxy required), but architecture
split into Gatekeeper + Restore layers so future Velocity migration is
mechanical. Trip-wires codified for when to reconsider.
Anti-drama policy locked in code (not config): no paid priority, no
hidden veteran tier, transparent ban appeals.
Bootstrap repo at git.s8n.ru/s8n/auth-limbo-v2 ready for M0 work.
Plugin renamed for clarity. Same functionality. Old name was tied
to racked.ru brand; new name describes what it does.
- Java package ru.racked.limbo -> ru.loginlimbo
- Main class RackedLimbo -> LoginLimbo
- Jar LoginLimbo-1.0.0.jar
- Command /loginlimbo (alias /llimbo)
- Permission loginlimbo.admin
- Log prefix [LoginLimbo]
Auth-limbo + login-restore fix for Paper 1.21+. Bypasses the
AuthMe `teleportOnLogin` race (PaperMC/Paper#4085) by listening at
MONITOR priority, reading the player's saved quit-location from
AuthMe's SQLite DB, pinning the destination chunk via
addPluginChunkTicket, then chaining getChunkAtAsyncUrgently and
teleportAsync.
Bundles a void auth_limbo world via custom ChunkGenerator so the
plugin removes the need for Multiverse-Core for offline-mode auth flows.
