I remain skeptical about the value of HFACS or similar categorization systems. They appear to place incidents / aspects into a box; this reactive approach might prejudice the underlying reasons - at risk from hindsight bias.
A more proactive alternative could be to remove ‘error’ from the investigation. The thread title is ‘human factors incidents’, not errors.
Hollnagel, Amalberti, and Woods provide interesting views of error; as I understand, one aspect replaces the concept of error with the variability of human performance.
Thus in seeking an understanding of recent events, you should consider how the individual or collective human performance varies, and what factors (the human factors) contribute to the variation, i.e. what is normal and what contributes to this normality.
Performance variation is normal; crossing an acceptable boundary of performance (analogous to error) depends on the contributions and interactions in the variability (functional resonance - Hollnagel), and where the boundary is placed – individually, collectively (professionally), or organizationally.
Have individuals been subject to external changes – contributions to variability, or have individual performances, ability, motivation, etc, changed, or slowly eroded.
Similarly look at the boundaries of acceptable performance; have these changed. Have operations slowly drifted towards the edge of acceptability – normalization of deviance, or have circumstances changed that require the boundaries to be redefined.
We tend to fix only what we find; thus ask why / what five times; the ‘fix’ might be deeply buried in the system.