Credibility requires honesty. A pattern that claims to be appropriate for every application is not a pattern; it is marketing. SAUI has real limits and stating them plainly is part of the argument, not a concession to it.
Real-time collaborative editing
When multiple users concurrently mutate the same document - a shared document editor, a collaborative whiteboard, a live code editor - the server-authoritative model requires conflict resolution that SAUI does not provide out of the box.
Concurrent writes to the same state require CRDTs (Conflict-free Replicated Data Types) or operational transforms to merge diverging edit histories. These are hard problems and the SAUI append-only event log does not solve them.
For single-user editing with server persistence - a CMS, a form, a document editor used by one person at a time - SAUI is entirely appropriate. The limit is specifically real-time multi-user concurrent mutation of shared state.
Canvas and creative tools
Tools like Figma, Photoshop, or a code editor have an interaction model where local state is inherent. Every mouse movement is not a server round-trip. Every keypress does not trigger a projection update. The latency of server authorisation is incompatible with the performance requirement.
These applications require a local state model with selective server persistence. The SAUI round-trip model is not appropriate here.
Offline-first PWAs
Applications that must function without a network connection require client-side state with sync-on-reconnect. A service worker can cache projections, but mutations made offline need to be queued and replayed when connectivity is restored.
SAUI is online-first. The gateway cannot authorise actions it cannot reach. For applications where offline operation is a primary requirement, not a nice-to-have, a client-side state model with server sync is appropriate.
Games and simulations
High-frequency local state - game physics, simulation state, real-time rendering - is a performance requirement. Server-authoritative game state exists (it is the standard cheat-prevention model in multiplayer games) but the implementation is specialised and the tooling is not general-purpose web infrastructure.
High-frequency financial data
Live tickers and order books can be delivered via SSE or WebSocket projection streams, and the SAUI model applies at the architecture level. At scale, however, this becomes an infrastructure problem: fan-out to thousands of connections receiving updates at high frequency requires careful design that goes beyond the pattern described here.
The SAUI model is sound for this use case. The implementation demands are significant.
Compliance, and what an architecture cannot certify
This site argues that a server-authoritative design answers several regulatory and audit requirements structurally rather than on paper. That argument has a firm ceiling, and it is worth stating here rather than only in the posts that make it. An architecture is not a compliance programme. GDPR, NIS2, the Cyber Resilience Act and SOC 2 all examine an organisation - its policies, its people, its records and its behaviour over time - and a system is only one of the things they look at.
The practical consequence is a wording rule, not a design rule. A system built to a standard's controls is architected to those controls. It is not compliant, and it is certainly not certified, until an auditor with the standing to say so has said it. Any vendor claim that an architecture alone delivers a certification is false, and that includes claims made on behalf of this one.
The honest boundary
If your application is a food ordering site, a banking dashboard, a booking system, a CMS, an admin panel, a SaaS product, or most of what is built for the web - SAUI is appropriate and the complexity of the SPA model is not justified.
If your application is a real-time collaborative editor, a creative tool, an offline-first mobile web app, or a high-frequency data display - evaluate carefully. SAUI may still apply partially, or may not be the right fit.