01 // Incident Baseline

Storage Allocation Escalated Three Months After Deprecation

In May 2026, the Apollo telemetry engine repository was formally decommissioned, marked read-only on primary remotes, and moved to tier-two cold storage. The entire volume measured exactly 18.4 GB across all commits, assets, and documentation. Routine weekly capacity audits should have reported a flat zero growth rate across the whole archive tree.

By early October, automated SAN monitoring flagged an acute quota warning on the underlying storage pool. The ostensibly frozen folder had swollen from 18.4 GB to 160.8 GB, claiming an additional 142.4 GB of high-speed block storage. The project manager confirmed that no developer commits had been registered since official sunsetting, presenting a perplexing anomaly for system administrators.

Experiencing Phantom Branch Growth in Deprecated Volumes?

Our differential snapshot engine detects isolated worktree growth and forgotten CI pipelines before they consume SAN quota limits.

Schedule Delta Diagnostic
02 // Differential Scan Audit

WizTree Comparison Uncovered 142 GB of Hidden Worktree Sprawl

Running a comparative WizTree snapshot delta between May 15 and October 10 illuminated the exact focal point of storage consumption. While the master and release branches remained completely static, a detached folder inside the hidden directory structure was expanding consistently every twenty-four hours.

  • Detached worktree directory .git/worktrees/nightly-perf failed to register in root directory quotas while continuously accumulating 4.2 GB daily dumps.
  • Automated cron scripts on secondary CI runner remained enabled, fetching the stale branch every midnight and compiling uncompressed debug binaries.
  • NTFS alternate data streams and build cache lockfiles prevented automatic garbage collection from cleaning up 38,400 orphaned object fragments.

The development team had disconnected the primary repository from the company dashboard during decommissioning. However, an auxiliary staging agent retained an active webhook and continued triggering automated benchmark jobs against a forgotten feature branch labeled experimental-matrix-cache.

We thought the project was dead and frozen in storage, but the local build runner treated the experimental branch as an active production target every single night.

— Gwen Bradley, Lead Storage Forensic Engineer
03 // Diagnostic Findings

File System Hash Diffs and Cache Bloat Breakdown

Differential hash auditing verified that core repository files were untouched. The entire volume inflation originated from unversioned temporary directories created by repeated compilation loops, which deposited diagnostic logs and intermediate shared libraries without cleanup scripts.

Storage Delta Inspection MetricDifferential Hash
Baseline Repository Size at Archival:18.4 GB (SHA256: 4e9a...81)
Peak Size at Discovery (3 Months Post-Archive):160.8 GB (+142.4 GB Delta)
Active Waste Identified in Detached Worktrees:139.1 GB across 38,412 files

Over three months, ninety-four consecutive night cycles had produced hundreds of redundant symbol files. Because the branch maintained its own git worktree outside the standard view of common file managers, administrators conducting visual spot-checks saw only the clean root directory.

04 // Resolution Protocol

Branch Pruning & Storage Reclamation Protocol

Remediation commenced by terminating the dormant cron schedule on the auxiliary runner host. The detached worktree link was formally removed using native git commands, followed by aggressive repository garbage collection with aggressive packfile repacking. In total, 142.4 GB of disk space was reclaimed immediately, restoring the archive back to its verified 18.4 GB baseline.