Observed Signal · Aug 18, 2026 · Technical Release · Source: DEV Community · Impact: 1/5 · Sentiment: Neutral
10-Minute Guard Reclaims 46GB Compressed Memory Leak
A developer discovered a macOS blind spot where compressed memory (CMPRS) is not counted in RSS, allowing the system daemon dasd to accumulate 46GB of compressed memory while reporting only 264MB RSS. The leak caused swap to balloon (≈37GB), which made Playwright operations time out and halted an automated social-posting environment. The author implemented mem-hog-guard.sh — a launchd-scheduled script running every 10 minutes that reads top's MEM (which includes CMPRS), converts units to MB, applies per-process thresholds and cooldown files, and restarts processes differently depending on whether they are system daemons or user launchd jobs. The script uses locking, DRY_RUN checks, PATH tightening for launchd, and sends aggregated Discord notifications. On the observed run the guard reclaimed dasd (logged at 02:32) and swap dropped to 4.1GB.
Practical, actionable operational guidance for macOS memory monitoring and automated recovery; useful to engineers but limited broad impact on the AdTech industry.
Track Apple Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- On 2026-08-09 the author observed dasd holding 46GB of compressed memory (CMPRS) while showing 264MB RSS, causing vm.swapusage to reach ~37GB.
- The author implemented mem-hog-guard.sh, run every 10 minutes via launchd StartCalendarInterval, which inspects top -stats pid,command,mem,cmprs and thresholds on the MEM column (includes CMPRS).
- Restart strategies differ: sudo killall for system daemons managed by launchd (dasd), and TERM → wait → KILL → launchctl kickstart for user launchd jobs (iii/com.lily.agentmemory).
- The guard uses per-process cooldown files, an shlock-based lockfile to prevent overlap, a DRY_RUN mode for verification, and sends aggregated Discord notifications after each scan.
- After deploying the guard the leak was reclaimed (log shows reclaim at 02:32) and swap usage was reported down to 4.1GB.
Connected Companies & Entities
5 Entities mapped“macOS `top` has a `cmprs` (compressed memory) column available via the `-stats` option....”
“A monitoring script runs every 10 minutes, restarts things when they cross a threshold, and a notification lands in Discord....”
“The Playwright instance managed by Claude Code looked like it was running....”
“Follow along: [Portfolio](https://bokuwalily.com/) · [X](https://x.com/bokuwalily) · [GitHub](https://github.com/bokuwalily)...”
“Follow along: [Portfolio](https://bokuwalily.com/) · [X](https://x.com/bokuwalily) · [GitHub](https://github.com/bokuwalily)...”
Ontology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
PMAD: Deleting malloc's slow path for deterministic allocation
An author published PMAD, a small open-source C99 allocator that removes the traditional “slow path” from malloc-style allocators to bound per-call latency. Benchmarks (macOS) show typical median allocations are nanoseconds while rare slow-path hits in system allocators produced single malloc calls as high as ~7 milliseconds. PMAD instead mmaps a fixed pool at startup, requires declared size classes, and implements alloc/free as O(1) free-list operations with no fallback, coalescing, growth, or locks. In head-to-head tests PMAD keeps the tail latency tight (P50→P99.9 spreads small) and under sustained fragmenting churn its worst case held around 40 µs versus the system allocator’s ~6.95 ms. PMAD is MIT-licensed, available on GitHub, targeted at special-purpose, per-core shared-nothing systems, and the published benchmarks are macOS-only.
Fixed 45% Idle CPU Leak in Retro Music Web App
A developer diagnosed and fixed a severe idle CPU leak in an interactive retro music web app that left an idle tab consuming ~45% CPU and steadily increasing memory. By isolating canvas layers, the author found uncontrolled requestAnimationFrame loops, object allocations inside 60 FPS loops, and stacked animation frames as the root causes. The fix used document.visibilitychange checks and pausing animation loops when audio was not playing; idle CPU dropped from 45% to 0.8% and background tabs suspended to 0%. The author also used Google AI (Gemini) during profiling and cites a GitHub repository for the project.
Robust Nightly LLM Log Drain Survives Sleep and Races
A technical walkthrough of a macOS shell script (vault-auto-ingest.sh) that nightly ingests Claude conversation logs into a local Obsidian vault, commits them to a private Git repo, and generates a daily brief. The author describes three real failure modes that drove the final design: sleep-related freezes when the Mac sleeps, double-execution races between scheduled and manual runs (noted on 2026-06-10), and a service-side timeout/outage that caused uniform slot failures on 2026-06-13. The resilient design uses four launchd slots (4:55, 8:20, 10:45, 12:15), caffeinate -i -s, atomic lockdir via mkdir, idempotent step markers for partial retry, explicit timeouts with --kill-after, preflight checks (network, TCC/Full Disk Access), and deterministic overwrites for LLM-generated numbers.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.