Content-Aware Fill fill generation repeatedly stalls during rendering. The render progress remains stuck until a timeline frame-change event occurs.

When generating a Content-Aware Fill layer, the rendering process repeatedly stalls and stops advancing on its own.

NOTE: I work with 8k monoscopic and work with reference frames every 1 second.

The fill generation begins normally and creates placeholder fill PNG files. However, progress eventually stops and remains stuck indefinitely. CPU usage remains active (approximately 8–15%), suggesting the process has not completely crashed.

The issue can be reliably reproduced and immediately resolved by causing a timeline frame change. Pressing Page Down, clicking a different frame in the timeline, or otherwise changing the current frame causes Content-Aware Fill to resume processing instantly. In many cases, a single Page Down key press causes the render progress to jump significantly or complete entirely.

The issue does not appear to be related to GPU acceleration or cache configuration. I have tested with:

• Mercury GPU Acceleration (CUDA)

• Mercury GPU Acceleration (OpenCL)

• Mercury Software Only

• Multi-Frame Rendering disabled

• Preferences reset

• Cache/database cleared

• Cache locations moved to a single SSD

The behavior remains identical in all cases.

The issue has been reproduced across multiple projects and with both short (1 second) and longer (4 second) fill regions.

An AutoHotkey script that automatically sends Page Down followed by Page Up every few seconds allows Content-Aware Fill to continue rendering unattended. This suggests Content-Aware Fill may be waiting for a timeline frame evaluation event before advancing to the next processing stage.

Expected behavior:
Content-Aware Fill should continue rendering without requiring any user interaction.

Actual behavior:
Content-Aware Fill stalls until a timeline frame change occurs.