| app-tracing-enforcement.md | Audit of reveried’s OTLP/Tempo tracing coverage: ~20 spans across 193 public functions, with only HTTP middleware instrumented. Recommends span additions at DB ops, event publishes, and child tasks. |
| engram-as-policy-substrate.md | Retrospective on evolving engram from a memory journal into an active policy substrate that shapes LLM routing, scoring, and consolidation decisions. |
| hotswap-listener-design.md | Design for a reusable hotswap-listener crate enabling zero-downtime binary restarts via SO_REUSEPORT + SIGUSR2; defers fd handoff and SCM_RIGHTS to later versions. |
| kernel-tracing.md | Survey of kernel-level process observability options for the reverie env-map; recommends procfs + /proc polling (zero root, ~5 ms/10 PIDs) over eBPF which is blocked on WSL2. |
| ldap-kerberos.md | Auth research for multi-host mesh scenarios; recommends doing nothing now (single-user localhost) and adopting mTLS via rustls+rcgen when remote GPU workers are introduced. |
| rust-monorepo-publishing.md | Survey of crates.io publishing workflows for Cargo workspaces, covering selective publishing, cargo-release, registry alternatives, and the recommended release pipeline for reverie. |
| ebbinghaus-stability-fsrs.md | FSRS v4 algorithm research (T50): full formulas for stability/retrievability/difficulty; concludes FSRS requires grade input reverie lacks — recommends retrievability-aware increment formula instead. |
| sqlite-backup-restore.md | SQLite backup options for T62: recommends rusqlite::backup online API (WAL-safe, non-blocking) for live backups; VACUUM INTO for cold snapshots; pre-dream auto-backup as the trigger. |
| dream-cycle-rate-limiting.md | Rate limiting design for T63: three-layer approach using Semaphore(1) + try_acquire_owned(), MissedTickBehavior::Skip, and minimum-interval guard in a single DreamGuard choke point. |