The customer will notice that Autoscaler in TAS v10.2.1+ and v6.0.18+ introduced a regression that scaled up stopped apps to their min configured instance count.
For example, app enabled autoscaler with Instance Limits Minimum:2 and Maximum:3. We scaled down app to one instance manually and stop it.
cf scale <app_name> -i 1
cf stop <app_name> The expected behavior is that app instance number won't be changed once stopped. However, in TAS v10.2.3 we will observe that Autoscaler scaled up stopped apps to its minimum configured instance count.
Getting events for app <app_name> in org <org_name> / space <space_name> as admin...
time event actor description
2026-03-25T10:17:39.00+0800 audit.app.process.scale instances: 2
2026-03-25T10:17:19.00+0800 audit.app.stop admin
2026-03-25T10:17:10.00+0800 audit.app.process.scale admin instances: 1
TAS v10.2.1+ and v6.0.18+
We introduced scale improvements for the Autoscaler in 10.2.1 and 6.0.18 which introduced a regression that scaled up stopped apps to their min configured instance count.
This has been fixed in the March patch for both 10.2 and 6.0: