I fell down a rabbit hole reading about Falcon 1's first three launches — the ones that failed before SpaceX finally made orbit on attempt #4 — and what got me is that none of them were some dramatic explosion-by-design. Each one traced back to a detail you'd never think twice about.

I ended up making hand-drawn animations for each mechanism because honestly I couldn't picture them from text alone. Quick version:

Flight 1 (2006): A single fuel-line nut corroded through and cracked. The nut was aluminum, the tube was stainless steel, and the launch site was a humid, salty island in the Pacific. Two different metals + saltwater = literally a tiny battery, and the aluminum slowly corroded away as the anode. Fuel leaked, the exhaust plume lit it, and it burned through the control lines in seconds. Fix: make the nut and tube the same metal. Done.

https://reddit.com/link/1ud24q6/video/exnqnjxhgx8h1/player

Flight 2 (2007): During stage separation the first stage lightly bumped the second stage's nozzle — barely a tap. The control system tried to correct the resulting tiny wobble, but the second-stage tank was half-empty, so the liquid oxygen started sloshing. The control frequency happened to match the slosh frequency, so every correction added energy instead of removing it. Same physics as soldiers marching in step collapsing a bridge. It "death-wobbled" for 3+ minutes until the engine sucked in pressurant gas instead of fuel and shut down.

https://reddit.com/link/1ud24q6/video/eixchwfkgx8h1/player

Flight 3 (2008): The new regeneratively-cooled engine (it runs kerosene through the nozzle walls to cool them) left a little residual kerosene in the channels after first-stage shutdown. In vacuum, that leftover fuel kept producing a whisper of thrust — enough that the spent first stage drifted forward and rear-ended the second stage. The kicker: it was undetectable on the ground because sea-level air pressure suppressed it. The fix was literally a few lines of code adding a ~3.5s wait. Musk later said one extra second of patience in the software would've saved the flight.

Three failures, ~$100M gone, and SpaceX was nearly broke. Then attempt #4 worked and the rest is history.

I'm not an aerospace person — I learned all of this while making the animations, so if I've mangled anything I'd genuinely love a correction from someone who actually knows engines. Just sharing here for fun. Full article: https://www.agenticbrew.ai/originals/the-butterfly-effect-behind-falcon-1-s-fatal-failures