L2a Forest Height (FP_FH__L2A) results look really noisy

[f2rn8gtg4h-dot]

After running the L2a processor on a TOMOGRAPHIC-phase stack over the Paracou test site (French Guiana), the resulting Forest Height product looks extremely speckled, with strong pixel-to-pixel variation across what should be relatively homogeneous forest. Values range from 0 to ~45.8 m, which is plausible in magnitude, but the spatial texture is dominated by salt-and-pepper noise rather than a smooth canopy-height field

I'd like to understand whether this is:

1. The expected native output of L2a FH (and smoothing/aggregation only happens at L2b), or
2. A processing-parameter issue I can fix in AUX_PP2_2A (e.g. resolution, upsampling factor, polarisation combination), or
3. A stack-quality / coregistration issue upstream.

Workflow (L1c → L2a)

• Inputs: 7× Sx_STA__1S stacks, TOM phase, S1 swath, Track 043 / Frame 150, cycles C01–C07.
• AUX_PP2_2A: default bio_aux_pp2_2a_*.xml.
• FNF: mounted via List_of_Cfg_Files.
• Run: L2A_P JobOrder_L2A_P.xml on Ubuntu (WSL). Completed without errors.

Question

Is this pixel-scale speckle the expected native L2a FH output (with smoothing happening at L2b), or is there an AUX_PP2_2A setting I should adjust? Or i used wronf FnF file? Happy to share the Job Order or logs if useful.