Autoscaling vs Manual Scaling: Availability‑Cost Trade‑off for Low‑Traffic Heroku Apps

Goal: Determine whether enabling Heroku autoscaling for a low‑traffic application yields lower monthly cost than running a single hobby dyno (with free dyno hours) when the app experiences intermittent requests.

Constraints: Free dyno hours provide ≈550 h/month but cause the dyno to sleep after 30 minutes of inactivity, adding latency on wake‑up; autoscaling can only be applied to hobby or professional dynos and bills each active dyno in full 24‑hour blocks, even if it runs for only a fraction of an hour. The potential sunset of free dyno hours adds uncertainty to the baseline cost.

Questions: Does autoscaling ever reduce the effective hourly cost compared to a continuously running hobby dyno? Under what request frequency or CPU‑usage patterns would autoscaling trigger enough to offset the per‑dyno block charge? How does the possible removal of free dyno hours shift the cost‑benefit balance between the two approaches?