# DevOps Weekly Roundup: Flux Turns 10, Argo CD v3.5 RC, and Platform Engineering Goes Mainstream - Link: https://thestackobserver.com/devops-weekly-roundup-flux-turns-10-argo-cd-v3-5-rc-and-platform-engineering-goes-mainstream/ - Published: 2026-07-08T11:07:42-04:00 - Author: Stackxx The DevOps landscape this week is defined by milestones and maturation. Flux turned ten years old, Argo CD shipped a major release candidate packed with security hardening, and platform engineering officially graduated from buzzword to baseline operating model. Here’s what you need to know. ## Flux Turns 10: A Decade of GitOps On July 7, 2016, Peter Bourgon made the initial commit to what would become the Flux project. Ten years later, that experiment has evolved into one of the most trusted continuous delivery systems in the Kubernetes ecosystem. The numbers tell a compelling story. Over the past decade, the Flux project has accumulated **1,076 contributors** across **44 repositories**, with **17,946 pull requests** and **7,474 issues** processed. On Flux 2 alone, the project has shipped** 210 releases** and seen **30.2 billion container image downloads**. But the maintainers are quick to point out that Flux is more than software. “GitOps cannot fully be embodied by our tools,” the project wrote in its anniversary post.“ Flux is the same. It is not purely software.” The project credits its longevity to a community built on technical excellence, collaboration, and a shared commitment to solving real operational problems. Today, Flux runs in environments most tools never touch: 5G towers, retail stores, cloud control planes, open science clusters, airplanes, tractors, satellites, and air-gapped networks. The maintainers attribute this trust to design decisions that prioritize security and reliability — avoiding binary execs, using pure Go SDKs, segmenting concerns across independent controllers, and relying on core Kubernetes API machinery. Looking forward, Flux is already adapting to the agentic AI era. The project recently published `fluxcd/agent-skills` to help coding agents use modern GitOps practices, and the Flux Operator MCP bridges runtime chaos to editor and agentic workflows. The next decade, the maintainers say, starts from a foundation that is “secure, open, and fast.” _Source: [Flux Blog](https://fluxcd.io/blog/2026/07/flux-turns-10/)_ ## Argo CD v3.5 Release Candidate Ships with Security and Scale Improvements The Argo CD community announced the v3.5 release candidate this week, and it is one of the most feature-rich updates in recent memory. The release promotes multiple alpha features to beta while adding new capabilities across authentication, Helm support, and supply chain security. **Impersonation moves to beta**, enabling Argo CD to assume different user identities for operations like viewing logs, deleting resources, and syncing applications. This is critical for multi-tenant environments where audit logging and RBAC need to reflect the actual user context. The **Source Hydrator** also advances to beta, now supporting separate repositories for dry and hydrated manifests. Teams can keep unhydrated source in one repository while storing hydrated output in another — a pattern that enables advanced multi- repository GitOps workflows. Source integrity checking on dry commits ensures only signed, trusted manifests are hydrated and deployed. ApplicationSets see significant stability improvements. They are now **fully stable in any namespace**, not just the Argo CD namespace, enabling independent multi-team GitOps workflows. A new **concurrency control** feature prevents ApplicationSet operations from overwhelming clusters or Git repositories with simultaneous requests. Other notable additions include **Helm 4 support**, **mTLS for repo-server** communications,** source integrity validation** for supply chain security, and **Azure Service Principal authentication** for Azure DevOps repositories. _Source: [Argo Project Blog](https://blog.argoproj.io/argo-cd-v3-5-release-candidate-02b1fbf7b419)_ ## HCP Terraform Infragraph Enters Limited Availability HashiCorp announced that **HCP Terraform powered by Infragraph** is now in limited availability for Standard and Premium customers using US instances. Infragraph addresses a problem that has grown acute as organizations scale: infrastructure data stored in silos across hybrid and multi-cloud estates. The platform connects directly to cloud and infrastructure sources — now including Microsoft Azure, GitHub, and Terraform Enterprise — to map resources, relationships, and state into a unified graph view. Teams can query this graph to understand ownership, identify risks, and surface unmanaged or orphaned resources. Early customer queries show practical value: viewing AWS and Azure resources connected to the same Terraform state file, auditing EC2 security rules and regions, and surfacing what’s not under management. The query interface supports low-code and JSON queries, with natural language-to-query support planned. HashiCorp also published guidance on using Terraform to discover and govern unmanaged Azure infrastructure — an increasingly common problem as AI workloads accelerate resource provisioning beyond standard workflows. _Source: [HashiCorp Blog](https://www.hashicorp.com/blog/hcp-terraform-powered-by-infragraph-limited-availability-launch)_ ## Backstage v1.52.0 Removes Immediate Stitching and Overhauls UI Spotify’s Backstage project shipped v1.52.0 with several breaking changes and significant performance improvements. The most impactful change: **immediate mode stitching has been removed**. All catalog stitching now uses deferred mode, which processes entities asynchronously via a worker queue. This removes a source of catalog latency and instability that had plagued large deployments. The catalog backend received substantial performance tuning. The entity list provider now fetches lists and counts as separate parallel queries, fixing a PostgreSQL CTE inlining issue that caused 10-40x slower catalog list views. Multi-column statistics on the search table and optimized batch queries with `= ANY(array)` instead of ` WHERE IN` round out the backend improvements. On the UI side, Backstage introduced **semantic color token families** (Accent, Announcement, Warning, Negative, Positive) and a new `NumberField` component. `Combobox` and `Select` components now support async collections with incremental loading and proper scroll pagination. A new experimental `@backstage/connections` package also debuted, intended to eventually replace the integrations concept with a more flexible system for external host connections and credentials. _Source: [Backstage GitHub](https://github.com/backstage/backstage/releases/tag/v1.52.0)_ ## OpenTofu Patches Critical Security Vulnerabilities The OpenTofu project released versions 1.12.3 and 1.11.10 this month to address multiple security advisories. The most serious vulnerability, tracked as **GHSA- q7j3-v8qv-22vq**, could allow arbitrary file reads during certain Git operations via a maliciously crafted URL. Additional fixes address SSH connection handling where errors could include unescaped input bytes, and high CPU consumption when running `tofu` against attacker-controlled servers. For users of OpenBao with state encryption, a crafted JWE could cause panics or hangs on compromised systems. The 1.12 series also fixed a race condition during `tofu login` signal handling and a panic when using ephemeral resources during tests. Users on affected versions should upgrade promptly. _Source: [OpenTofu GitHub](https://github.com/opentofu/opentofu/releases)_ ## Platform Engineering Is Now the Default, Not the Exception Multiple industry voices confirmed this week what practitioners have been feeling for months: **platform engineering has graduated from buzzword to operating model**. A report from Platform Engineering noted that by 2026, what began as a reaction to cloud-native cognitive overload has become, in Gartner’s framing, the standard approach to software delivery at scale. The shift is showing up in hiring patterns, organizational structures, and tooling investments. Modern DevOps roles now routinely span AWS, Azure, and Google Cloud simultaneously. Internal developer platforms are no longer experimental — they are the expected infrastructure for engineering teams. But maturity brings new challenges. A Forbes Tech Council article this week warned that **DevOps initiatives lose their way when platform teams become gates instead of self-service enablers**. The razor-thin line between “uncontrolled sprawl” and“ over-engineered platform” is the central tension IT leaders now navigate. CircleCI’s recent post on agentic validation reinforced the theme: as AI-generated code accelerates, the infrastructure for validating that code must evolve. Traditional CI/CD pipelines weren’t designed for agentic workflows, and the gap is where the next wave of DevOps tooling will compete. _Sources: [Platform Engineering](https://platformengineering.com/features/what-is-platform-engineering-inside-the-discipline-reshaping-modern-software-delivery/), [Forbes Tech Council](https://www.forbes.com/councils/forbestechcouncil/2026/07/01/clear-warning-signs-a-devops-initiative-is-losing-its-way/)_ ## Also Notable **Tekton Pipeline** shipped patch releases v1.6.5 and v1.9.6 with Go bumps for CVE remediation and signed attestations via Sigstore Rekor. **GitHub** expanded secret scanning validity checks to include additional metadata— owner details, creation and expiry dates, and project or organization context — making tririage and remediation more actionable across alerts, webhooks, and the REST API. **Dynatrace and NVIDIA** announced a partnership integrating Dynatrace observability with NVIDIA AI-Q, aiming to make agentic AI deployments smarter and safer through unified telemetry. **CircleCI** rebuilt its CLI from scratch and published new guides on wiring sidecars into agent hooks to prevent broken code from reaching CI in the first place.