The hardest part here is pushing a light gantry into thick Hardox, and the maker went straight at it with an insert drill. The feed override was left sitting at 250 percent, the tool bit in, screeched, stalled, and the whole thing ended in a violent crash that snapped the inserts and the drill. The machine halted after impact, leaving the broken tool hanging over the cut. That is a costly way to learn a limit, but learning limits is exactly the job the maker has been given: finding what the machine does reliably and which work to turn away.
The setup was deliberately rough. The plate sat on a wooden riser on the grooved bed, and the maker tests with crude workholding on purpose, so that any process proven this way can later be promised to paying customers. The shop cuts plate for a living, competing with hand-ground parts, and the work only needs 20 thou tolerance at worst. The machine itself has under one hour of running time, and the maker refuses to spend more on it until jobs pay for upgrades.
What stands out is the diagnosis. The maker reports the whole beam and gantry swayed during the cut, not the spindle bearings, and points to a nasty resonance between 200 and 600 rpm that makes drilling impossible. Pecking was in use, though the maker says it did not break the tool. Earlier attempts with solid carbide, interchangeable tip drills and annular cutters all failed; only HSS drills cut clean holes, and those are far too slow in Hardox.
The way forward is already mapped: plasma holes from underneath, then mill the profile to size, an approach that has already given great results. The open question is the practical size cutoff for parts this machine can handle, which testing still has to answer.