Evaluate Tauri against Electron after the reliability foundation is complete

Outcome

A shell migration should follow measured product benefits. Rewriting working domain and agent code is not a prerequisite for reliability.

Priority: P3 · Deferred experiment

Build a limited Tauri host around the same desktop-independent TypeScript engine and representative React screens.
Retain Electron as the shipping baseline during the comparison.
Measure startup time, idle and active memory, distribution size, and behavior under the same workload and hardware.
Verify embedded preview rendering, cookies/authentication isolation, credentials, cancellation, and recovery on macOS, Windows, and Omarchy.
Write an architecture decision with measured benefits, regressions, maintenance cost, and a keep/switch recommendation.

Measurements are reproducible and include versions, hardware, workload, and repeated runs.
The comparison includes preview and platform behavior, not just a blank window.
Any Node sidecar and native build complexity are counted in the result.
No production migration occurs merely because the prototype compiles.
The experiment starts after the reliability acceptance gate; no full Rust or Go rewrite is included.

Use official framework documentation and attach measured results. A decision to retain Electron is a valid completed outcome.

Extract a desktop-independent Studio engine and diagnostic CLI #3 — Extract a desktop-independent Studio engine and diagnostic CLI
Establish the reliability sprint acceptance and fault-injection suite #13 — Establish the reliability sprint acceptance and fault-injection suite