The benchy is a lie. Not the model, the model is fine, it's a clever little stress test of a boat. The lie is the implication that printing one means you're done calibrating. My first benchy came out passable and I believed, briefly, that I understood this machine. The bin full of casualties since then has corrected me.
I want to keep a record of the early failures while they're fresh, partly because the next person setting up the same printer will hit the same ones, and partly because I'll forget and have to relearn them.
the spaghetti
First real failure was the classic: a tall print that let go of the bed about a third of the way up, after which the nozzle spent two hours extruding a nest of filament into thin air. I came back to a bird's nest of PLA and a printer that had, with great diligence, printed nothing whatsoever.
The cause was adhesion, obviously, but the lesson was supervision. A failed first layer is cheap. A failure at layer 200 has cost you two hours and a spool's worth of plastic before you notice. I now watch the first few layers go down every single time, and I've stopped trusting "it'll be fine" on anything tall.
the under-extrusion
Second one was subtler. Prints that looked right but felt wrong: gaps between the top layers, walls you could see daylight through, a part that snapped when it should have flexed. This isn't a dramatic failure, it's a quiet one, and it took me a couple of ruined parts to spot the pattern.
It was the filament. A roll that had been open on the shelf for a few weeks had pulled in enough moisture to stutter on extrusion, popping and hissing at the nozzle. A few hours in a cheap food dehydrator and the same roll printed clean. I now keep opened spools in a box with desiccant and treat "it was fine last month" with suspicion.
the warping
Third was geometry, not process. A wide flat part curled its corners up off the bed as it cooled, lifting the edges into a gentle banana. PLA does this less than ABS but it still does it, and a big flat footprint is exactly where it shows.
A brim fixed it, mostly. A wider first-layer footprint to hold the corners down while the centre cooled, peeled off afterwards with a thumbnail. The deeper lesson is that the print and the cooling are fighting each other, and on big flat parts the cooling wins unless you give it something to hold onto.
what the failures were for
None of these were the printer being broken. Every one was me not yet knowing the machine: how it likes its bed, how it feels about damp filament, what it does to a wide part as it cools. The benchy proves the motors move and the hotend melts. The failures are where you actually learn the thing. I'm keeping the bin of casualties on the shelf for now, as a reminder that "it printed once" and "I understand this printer" are very different claims.