Black Friday and similar peak shopping events subject e-commerce platforms to transaction volumes many times their normal load, compressed into hours, with the financial stakes of failure measured in millions of dollars per hour of downtime. Building systems that gracefully absorb these spikes without collapsing requires deliberate performance architecture rather than simply adding servers. This paper examines the architectural patterns that enable e-commerce platforms to handle extreme transaction spikes, analyzing load shedding, caching hierarchies, asynchronous processing, database scaling strategies, and graceful degradation. We designed a reference e-commerce architecture incorporating these patterns and subjected it to simulated load testing scaling from baseline to 50x peak load, comparing pattern combinations against a conventionally scaled baseline. Results show that the pattern-based architecture sustained 50x load with 99.4% transaction success rate and median latency under 400 milliseconds, while the conventionally scaled baseline collapsed at approximately 12x load with cascading failures. Multi-tier caching absorbed 78% of read traffic before it reached the database, and asynchronous order processing decoupled the customer-facing checkout from downstream fulfilment, preventing backend slowness from blocking sales. Graceful degradation strategies maintained core purchasing functionality even when non-essential features were shed under extreme load. The paper discusses the architectural trade-offs, the critical importance of load testing at realistic scale, and the organizational preparation required to survive peak events, offering a practical playbook for e-commerce performance engineering...