An architect is designing an API integration where Claude processes large document analysis requests that take 45–90 seconds to complete. The client applications cannot maintain long HTTP connections reliably. Which delivery pattern should the architect implement?

C is correct because webhook-based async delivery decouples request submission from result delivery entirely — the client fire-and-forgets the request, receives an immediate acknowledgment with a job ID, and receives results via callback when ready. This pattern is resilient to client connection instability and network interruptions during processing. B is wrong because polling wastes bandwidth and server resources across 45–90 second processing windows (18+ poll cycles at 5-second intervals per request), and adds latency equal to the poll interval after completion. A is wrong because SSE requires the client to maintain an open connection for the full processing duration — precisely the reliability problem the architect is trying to solve. D is wrong because extending client timeouts is a fragile workaround that fails on mobile networks, serverless functions with execution time limits, and environments where proxies impose their own timeout limits.