SCTransform + RunPCA warning: Unscaled features due to scale.data/var.features mismatch - mitigation strategies?

[Horrrse]

After running SCTransform followed by RunPCA, I encountered this warning:
Warning: The following 2489 features requested have not been scaled (running reduction without them)

This aligns with multiple reports #8880 that the issue occurs when scale.data contains significantly fewer features than var.features.

A potential fix is reducing min_cells to retain more features in scaling.

Is this adjustment (min_cells reduction) a recommended practice in the community?
Could this introduce any negative impacts (e.g., noise amplification, batch effect sensitivity)?

[Horrrse]

My major concern is that when these 2,489 unscaled features were excluded from PCA, the resulting dimensionality reduction and clustering produced abnormal results - cells appear poorly separated/mixed together.