Container stopped for high CPU at 15.88% — what is the actual threshold?

Free plan, node FREE EU RO 22, support tag 0c06a76c. My Python container was stopped twice with "stopped due to high CPU usage". First time: Average CPU 31.84%. Second time: Average CPU 15.88%. My panel shows a CPU limit of 35%, so the second stop was well under it. My process measures its own usage at 0.31% — 49 seconds of CPU over 4.4 hours of uptime. It sleeps most of the time and makes a handful of HTTP requests per hour. How is "Average CPU" calculated, and what is the actual threshold? I would like to stay within it but I cannot tell what the number means.