We do Class III implant components (spinal screws and plates) so I can add some specifics on what the FDA auditors actually look at.
The CPK 1.67 requirement is the minimum — our customer specs call for 2.0 on critical-to-function dimensions. The servo press force data is half the story. The other half is correlating force curve features to actual dimensional outcomes. We ran a DOE with 5 force levels x 3 speeds x 2 lube conditions and built a regression model. Now we can predict dimensional drift from the force curve alone without pulling parts for measurement. That model is what got us through our last FDA audit — they loved that we could demonstrate real-time process control, not just post-hoc inspection.
On traceability — we went beyond stroke counter sync. Each part gets a 2D Data Matrix laser-marked within 3 seconds of forming (inline station after the press). The mark encodes: press ID, die ID, stroke number, date/time, and a hash of the force curve peak/integral. If a surgeon reports an issue with an implant 10 years from now, we can pull the exact forming conditions. Cost us about $45K for the marking station but it's non-negotiable for Class III.
Biggest lesson: your validation protocol needs to include the press thermal state. We failed an OQ run because we started stamping immediately after a die change. The press frame hadn't reached thermal equilibrium and BDC was drifting for the first 200 parts. Now our SOP requires 15 minutes of dry cycling before any validation run starts.