Crash Loops, Idempotency Keys, and Query Log Serialization

Bug Fix

Fix: Crash Loop Segv on Container Startup

Kubernetes containers were crashing and burning due to a segmentation fault at startup. This merged PR apparently fixes the memory violation, so you can finally stop checking the k8s logs with your morning coffee.

Precise CPU Metrics Expose Hidden Scheduler Spikes

The scheduler was burying CPU spikes under generic "Load" metrics, making impossible to track. Now it exposes real user and system time, which is vital if you don't want to wake up to a 200ms latency dashboard without warning.

Fix: Deep Write Queue Bloat on Reconnect

This fixes the buffer bloat issue causing jitter during reconnection storms. Critical if you rely on this for users with bad hotel Wi-Fi, which is basically everyone.

Architecture

Query Log Serialization Resolves Write Chaos

Linearizing the query log creates a dependency bottleneck, but it fixes chaotic write patterns. If you are running high-throughput clusters, watch for this merge, it fundamentally changes how load is handled.

Design

RFC: Idempotency Keys for Critical Actions

Idempotency is boring until your network flakes and you execute a payment twice. Discussing this now suggests the core API is grappling its hardest scaling problem: exactly-once delivery.

Editorial

Dep Resolver Deep Dive: The 'Chatty' Latency Bug

Something is definitely wrong with the resolver if we need over 20 comments to agree on the fix. This is a 'look at the tracing' level issue, likely masking slowness that feels like ghost lag in your charts.

Operations

Strict Mode: Crashing on Missing Build Locales

A build system that silently decides "not to fail" on missing dependencies is just asking for a midnight incident. They are finally cleaning this up so your CI actually fails when the base image is broken.