Testing Stock API Recovery After Disconnects, Not Just Streaming Speed

A stock API may reconnect quickly yet leave an application with incomplete subscriptions or a gap in its event history, so evaluators should measure when data becomes usable again rather than when a socket opens. A fresh snapshot can restore current state, but only a documented replay or backfill mechanism can establish event continuity. The recommended evaluation defines two separate acceptance contracts, one for state restoration and one for history, before any testing begins. Recovery should be measured across distinct intervals, including detection delay, transport recovery, and usable-data recovery, using a monotonic local clock. Runs that never recover within the observation window should be counted as failures rather than dropped from latency averages.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
Discussion (0)
Log in to join the discussion and vote.
Log in