Anaconda Navigator 2.4.1 fix resolves tab‑close leak but bulk‑update memory growth remains unexplained

Analysis of Memory Growth vs. Tab-Closure Leak

The residual memory increase observed during bulk package updates in Anaconda Navigator ≥2.4.1 is likely distinct from the tab-closure leak fixed in version 2.4.1. While the previous leak was a failure to destroy the view-widget (requiring deleteLater()), bulk updates occur within an existing, active tab. This means the primary widget remains alive, shifting the source of memory growth from widget retention to resource accumulation within the QtWebEngine process.

Likely Resource Culprits

Because QtWebEngine is based on Chromium, it manages memory across a multi-process architecture that is largely opaque to standard Python profilers. During bulk updates, the following Qt objects and mechanisms typically retain memory:

• QtWebEngineProfile Cache: The browser profile caches network responses, DOM fragments, and JavaScript state to accelerate subsequent renders. Bulk updates involve heavy data ingestion, which expands this cache.
• QNetworkReply Objects: If network requests for package metadata are not explicitly deleted after completion, they can accumulate in the heap.
• Chromium Render Process: The underlying browser process may increase its Resident Set Size (RSS) to handle larger JSON payloads or complex table renders without immediately returning that memory to the OS.

Verification Steps

To determine if this growth is a leak or intended caching, perform the following scoped tests:

1. Linearity Test: Compare the memory growth of a 10-package update versus a 100-package update. Linear growth suggests a leak (e.g., QNetworkReply retention); logarithmic or plateauing growth suggests standard Chromium caching.
2. Profile Clearance: In a debug environment, manually trigger a clear of the QtWebEngineProfile. If the memory returns to the baseline, the growth is attributed to the profile cache rather than a leaked object.
3. Object Tracking: Use a tool like GammaRay to monitor the count of active QObject instances. If the number of QWebEngineView or QNetworkReply objects increases and never decreases after the update completes, a leak is confirmed.

Diagnostic Requirement

To refine this recommendation: Does the memory usage plateau after a specific number of updates, or does it continue to climb indefinitely regardless of the package count?