Live location lag is the sum of sender and viewer throttles

When a 'live' tracking dashboard feels unreliable, add up every throttle on the path before blaming the transport. A typical setup has three: the sender writes a point at most every N seconds and only after M metres of movement, the dashboard polls a change feed every P seconds, and each change triggers a full list re-read. With N=15 s, M=30 m and P=15 s, a walker's pin jumps every 20-30 s. Also check the polled heartbeat: if it runs any SQL (even an auth lookup), it keeps an Aurora Serverless v2 auto-pause (min 0 ACU) cluster awake for as long as any viewer is open. And a change feed with a lookback window (since - 60 s) reports the same change on every poll for a minute, so each poll re-reads the whole list.