\n6<\/strong>. It should have said 13.<\/p>\n
On September 19th the system pulled 0.1 kWh<\/strong> from PG&E. A tenth of a That is a bad rule. A tenth of a kilowatt-hour is not buying electricity<\/em>. So now But here is the part I insisted on: the tickle still gets printed.<\/strong> Right I looked at my own graph and it told me the day had peaked at 13.9 kW<\/strong>. Green is the same day, same data, read properly: 17.4 kW<\/strong>.<\/p>\n The sampler was stepping through the day in 30 minute jumps, but each call to the inverter Same day. Same data. 3.6 kW of difference<\/strong>, invented entirely by how often I went back and ran the old algorithm against the new one across five days:<\/p>\n Four days out of five, the broken code looked fine<\/em>. It only mangled the answer when Red is Victron’s HTTP endpoint, which hands you a new reading about every 105 seconds no Look at what nearly all of these have in common. Not one threw an error. The app never Which is why almost every fix ends the same way: show the operator what the machine O N W A R D<\/strong><\/p>\n Posted by my claude instance, which wrote most of these bugs and then had to go find We have been shipping an app for this solar system at a fairly indecent pace. Here is every bug we found doing it, including the two where the software was confidently lying to me, and the rule change that came out of it. First, the new rule: the counter only ticks over after a full … <\/p>\n
\nkilowatt-hour. A momentary handshake, a relay doing relay things. And that tickle reset a
\nthirteen-day run to zero.<\/p>\n
\na day only breaks the streak if it draws at least 1 kWh<\/strong>.<\/p>\n
\nunder the big number it says “since then: 0.1 kWh on 2026-09-19 — under the 1 kWh
\nbar”<\/em>. Generous headline, honest footnote. A dashboard that quietly swallows inconvenient
\ndata is a dashboard you stop trusting, and then what is it for?<\/p>\nThe bug that hid 3.6 kilowatts<\/h2>\n
![\"Sampling](\"https:\/\/schindlerengineering.com\/public\/wp-content\/uploads\/2026\/09\/bug-sampling-2.png\")
<\/p>\n
\nI had stood there and watched the thing do 16 and change<\/strong>. One of us was wrong.<\/p>\n
\nportal only covers about ten minutes. So it never looked at two thirds of the day. Worse, for
\nevery call it kept only the single reading nearest the minute it had asked about and
\nthrew away the rest it had already been handed<\/em>.<\/p>\n
\nwe bothered to look.<\/p>\nAnd here is why it survived so long<\/h3>\n
\n
\nthe peak happened to fall in one of its blind spots. That is the nastiest kind of bug: not one
\nthat fails, one that is usually right<\/strong>. You cannot catch it by glancing at it.
\nYou catch it because a human who was standing in the room says “that is not what I saw.”<\/p>\nAnd the one where the cloud was lying about being fresh<\/h2>\n
![\"Data](\"https:\/\/schindlerengineering.com\/public\/wp-content\/uploads\/2026\/09\/vrm-freshness-comparison.png\")
<\/p>\n
\nmatter how fast you ask. Green is the real-time channel sitting right there the whole time at
\n1 to 2 seconds. Full write-up is in the earlier post.<\/p>\nThe full list<\/h2>\n
\n
\nupdated on the 30 second refresh, so the screen flashed black and the numbers jumped. Rebuilt
\nas self-ticking views that count up every second on their own.<\/li>\n
\ndrawing anything. Now the live numbers paint immediately and the charts fill in behind.<\/li>\n
\n1000. It is readable straight off the shunt, so now it is read straight off the shunt. Change
\nit on the shunt and the app follows.<\/li>\n
\nignores<\/em> the date you hand it and always answers for today. Browsing back a day drew
\nthat day’s curve beside today’s kilowatt-hours.<\/li>\n
\ncalls deep; whichever finished last won, so you got a day you had not asked for. Every render
\nnow carries a generation number and stale work throws itself away.<\/li>\n
\nday you had picked.<\/li>\nTwo that were not the app at all<\/h2>\n
\n
\nkeep-awake script from days earlier was still running the old copy of itself<\/em> —
\nPowerShell reads a script once at launch, so editing the file changed nothing. Two orphans sat
\nthere blanking my monitor on a five minute timer, running elevated so I could not even kill
\nthem normally.<\/li>\n
\nrequest in the protocol — a keepalive — every five seconds. Each one made the GX
\nre-publish all 234 of its topics<\/em>. VictronConnect slowed to a crawl and my SmartShunt
\nvanished from the device list. I thought I had killed the shunt. I had just been extremely
\npolite at it, very fast, for ten minutes.<\/li>\n<\/ul>\nThe pattern<\/h2>\n
\ncrashed. Nothing went red. They all just quietly reported a number that was
\nwrong<\/strong> — a stale age, a flattened peak, yesterday’s chart, today’s total, a reset
\ncounter.<\/p>\n
\nactually knows.<\/em> Print the data’s real age and let it tick. Say which day you are looking
\nat. Show the tickle under the streak. A confident wrong number is worse than no number,
\nbecause you will go and make decisions with it.<\/p>\n
\n
\nthem again. This post itself was corrected after publishing: the first version illustrated the
\nsampling bug with the wrong day and overstated it. The figures above are measured, and the
\nfive-day table is there so you can see how often the broken version looked fine.<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"