<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/npm-s-validation-pipeline-stalls-everyone-hits-saturation-limits-on-their-nics-kljx8dkcv" -->

---
title: npm's validation pipeline stalls, everyone hits...
description: npm is stuck in a broken state where publishes take 25+ minutes to validate and there's no workaround except waiting or skipping the audit step. A...
canonical: https://daily.dev/posts/npm-s-validation-pipeline-stalls-everyone-hits-saturation-limits-on-their-nics-kljx8dkcv
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: npm's validation pipeline stalls, everyone hits saturation limits on their NICs | daily.dev
og:description: npm is stuck in a broken state where publishes take 25+ minutes to validate and there's no workaround except waiting or skipping the audit step. A...
og:url: https://daily.dev/posts/npm-s-validation-pipeline-stalls-everyone-hits-saturation-limits-on-their-nics-kljx8dkcv
og:image: https://api.daily.dev/og/posts/KlJX8dkCV.png
og:image:alt: npm's validation pipeline stalls, everyone hits saturation limits on their NICs
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# npm's validation pipeline stalls, everyone hits saturation limits on their NICs

**[Backend Digest](https://daily.dev/sources/backend_digest)** · 4 min read · 1 upvotes · 0 comments

## Summary

npm is stuck in a broken state where publishes take 25+ minutes to validate and there's no workaround except waiting or skipping the audit step. A load-balancer ordering bug shows why your zero-downtime deploy might be silently doubling p99 latency. DuckDB quietly removed the last reason you needed a Spark cluster for lakehouse work. And a deep dive into 100GbE tuning found IOMMU quietly eating your bandwidth the whole time.

## Content

**TLDR:** npm is stuck in a broken state where publishes take 25+ minutes to validate and there's no workaround except waiting or skipping the audit step. A load-balancer ordering bug shows why your zero-downtime deploy might be silently doubling p99 latency. DuckDB quietly removed the last reason you needed a Spark cluster for lakehouse work. And a deep dive into 100GbE tuning found IOMMU quietly eating your bandwidth the whole time.

---

## npm's publish pipeline is stuck, and there's no way to force a retry

New packages are taking 25-30+ minutes to become available after publishing, and some are stuck in 'Validating' indefinitely with no error message. Armin Ronacher is reconsidering npm as pip's distribution channel over this, and Devon Govett got hit with downstream installs failing on a dependency that technically published but isn't reachable yet. The only workaround found so far is `npm i --no-audit`, which suggests the audit service is the bottleneck. No official status update from npm as of writing. [Read more](https://daily.dev/posts/mBsLTgflx)

## Your rolling deploy's zero-error dashboard might be hiding doubled p99 latency

nginx's default proxy_next_upstream retry silently re-runs requests whose backend connection dropped mid-deploy instead of surfacing an error, which nearly doubles p99 while your error rate stays at zero. A SIGTERM drain handler helps but doesn't fully fix it, because the load balancer can still route to an instance that already stopped accepting connections. The actual fix is ordering: pull the instance from the LB first, wait for that to propagate, then send SIGTERM — which is exactly what a Kubernetes preStop sleep hook is for. [Read more](https://daily.dev/posts/kun77SDUn)

## DuckDB 1.5.3 kills the last reason you needed Spark for local lakehouse work

DuckDB's May 2026 release added full Iceberg write support — MERGE INTO, ALTER TABLE, partition transforms, Iceberg V3 — closing the gap that used to force local workflows onto a Spark cluster. Paired with dbt Core's free unlimited-seat license and Iceberg's own maturation (Polaris now a top-level Apache project), you can run a full enterprise lakehouse stack on a laptop for free. Real gaps remain around concurrent writes and production observability, so this replaces learning environments, not production Snowflake or Databricks clusters. [Read more](https://daily.dev/posts/ref3HgDd3)

## 100GbE saturation comes down to sysctl tuning and one BIOS setting nobody checks

Getting a 100GbE server to actually hit its advertised bandwidth required BBR congestion control, fq qdisc, and tuned tcp_mem/wmem/rmem, plus NIC-level tuning of channels, ring buffers, and interrupt coalescing. The single biggest win was disabling IOMMU in BIOS, which was silently destroying performance the whole time. Go's built-in HTTP server also beat an nginx reverse-proxy setup by skipping extra kernel-userspace copies, and HTTP/2 outperformed HTTP/3/QUIC for raw throughput in this setup. [Read more](https://daily.dev/posts/AzD9fUIZ0)

---

## Also notable

- **LINE Plus's SRE Observer cuts MTTR in half:** The system correlates alerts via trace topology and LLM-judged similarity, then runs a hypothesis-driven root-cause agent against five failure categories, reporting 85-95% alert-noise reduction and a 50% cut in mean time to identify incidents. [Read more](https://daily.dev/posts/fb8F1upW2)
- **Cribl's credit pricing is pushing teams toward OTel Collector, Vector, or Last9:** A 2TB/day team on Cribl's cloud workers is looking at roughly $230,000/year at $0.32/GB, driving evaluation of open-source pipelines (OpenTelemetry Collector, Vector, Fluent Bit) and full-stack alternatives like Last9 that fold pipeline shaping into storage and alerting. [Read more](https://daily.dev/posts/qFQzdZiGx)
- **A parallel Python 2/3 deploy strategy turned a stalled migration into a routine release:** A team downgraded its modern Python 3 branch to 3.6 and ran it on old infrastructure alongside the legacy Python 2 version until the codebases matched, making the eventual cloud cutover a pure infrastructure change with zero code-behavior ambiguity. [Read more](https://daily.dev/posts/O9ipdzIob)
- **trynix runs any of 310,000+ nixpkgs versions in-browser with no install:** The tool boots a WebAssembly Linux kernel and streams store-path closures from any CORS-enabled Nix binary cache, letting you run archived packages like Python 3.6.2 from 2017 directly from a URL. [Read more](https://daily.dev/posts/w1YBVIi2L)

## Similar posts on daily.dev

- [Package Managers Need to Cool Down](https://daily.dev/posts/package-managers-need-to-cool-down-ee6vqq1u6) · Lobsters · 0 upvotes · 0 comments

---

Tags: [#kubernetes](https://daily.dev/tags/kubernetes), [#npm](https://daily.dev/tags/npm), [#nginx](https://daily.dev/tags/nginx), [#duckdb](https://daily.dev/tags/duckdb)

[View this post on daily.dev](https://daily.dev/posts/npm-s-validation-pipeline-stalls-everyone-hits-saturation-limits-on-their-nics-kljx8dkcv)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"DiscussionForumPosting","mainEntityOfPage":"https://daily.dev/posts/npm-s-validation-pipeline-stalls-everyone-hits-saturation-limits-on-their-nics-kljx8dkcv","headline":"npm's validation pipeline stalls, everyone hits saturation limits on their NICs","text":"npm is stuck in a broken state where publishes take 25+ minutes to validate and there's no workaround except waiting or skipping the audit step. A load-balancer ordering bug shows why your zero-downtime deploy might be silently doubling p99 latency. DuckDB quietly removed the last reason you needed a Spark cluster for lakehouse work. And a deep dive into 100GbE tuning found IOMMU quietly eating your bandwidth the whole time.","url":"https://daily.dev/posts/npm-s-validation-pipeline-stalls-everyone-hits-saturation-limits-on-their-nics-kljx8dkcv","datePublished":"2026-09-07T04:18:49.100Z","dateModified":"2026-09-13T21:12:11.269Z","author":{"@type":"Organization","name":"Backend Digest","logo":"https://media.daily.dev/image/upload/s--9uCSrsAb--/f_auto,q_auto/v1773839398/logos/backend_digest?_a=BAMAMiiu0","url":"https://daily.dev/sources/backend_digest"},"interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"isPartOf":{"@type":"WebPage","url":"https://daily.dev/sources/backend_digest","name":"Backend Digest"}}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Backend Digest","item":"https://daily.dev/sources/backend_digest"},{"@type":"ListItem","position":3,"name":"npm's validation pipeline stalls, everyone hits saturation limits on their NICs"}]}
```

