The Friday bug
Shows up at the end of the week, never when anyone is watching, and leaves before Monday.
Service
Cold case review
Some bugs never get solved. Teams just get used to them. We reopen the ones you've learned to live with and close them properly, with evidence.
$4,900 fixed · two weeks · a report, a fix and a much quieter on-call rota
What we take on
If it has been open longer than the intern has been employed, it's probably ours.
The memory leak with a schedule
Everyone restarts the service at three in the morning. Nobody remembers why. It has a cron job now.
The flaky test
Fails one run in twenty. Retried until green, then forgotten, until the release day it isn't.
The ghost complaint
Customers report it, support confirms it, engineering can't reproduce it. Everyone is telling the truth.
The page that was always slow
So slow for so long it became the baseline. We bring a stopwatch and no respect for tradition.
The workaround with a wiki page
A fix so temporary it now has documentation, an owner and a birthday.
Prime suspect
A decision nobody remembers making
Usually not the code itself, but a choice kept alive because everyone was too busy to question it.
Last seen: Friday, 17:59
Witness: the intern. Says nothing. Knows everything.
How it runs
1234567891011121314
Day 1 · Digging up the file
Bring everything: logs, tickets, the chat thread from two years ago and the colleague who says it works on their machine. Especially them.
Days 2–8 · Reconstructing the scene
We reproduce the problem before we touch it. Traces, timelines, a few uncomfortable questions for the codebase. Half the job, and it shows on the calendar.
Days 9–12 · The fix
One change at a time, each with a test, so everyone knows which one mattered. No heroic rewrites.
Days 13–14 · Closing the file
A short report, a walkthrough call and the next three suspects, in case you want to keep going.
Shaded days are weekends. The detective rests; cold cases, famously, don't.
What you get
A case file, not a slide deck. It says what happened, why it kept happening and what stops it from coming back, in plain language and in that order.
- The cause, with the evidence that proves it
- The fix, merged and covered by tests
- A timeline of how the bug survived this long
- Three suspects worth a look next
Report · Case 0311
Page 1 of 3
The 3 a.m. restart
- Symptom
- Service slows down every night; restarted by cron at 03:00 since 2022.
- Cause
- A cache that never forgot anything.
- Evidence
- Memory profile over 72 hours, attached as exhibit A.
- Fix
- Expiry policy plus the test that would have caught it.
- Status
- Cron job retired with honours.
Questions
Transcript · first call with a client, abridged
Q.Why do you call them cold cases?
A.Because the trail has gone cold, and so has everyone's enthusiasm for it. We bring a fresh pair of eyes and a thermos.
Q.What if you don't find it?
A.Then you don't pay the second half, and you keep everything we ruled out, which is often half the answer. The detective takes it personally, at no extra charge.
Q.Can you fix it in a week?
A.We can usually find it in a week. Fixing it properly depends on how much has been built on top of it since.
Q.Will the report blame someone?
A.It names causes, not people. Bugs this old are rarely one person's fault; they're everyone's habit.
Slowburn · Bug investigations
Form SB-7 · Case intake
Reopen a case
Tell us what keeps happening. We'll reply within a working day, slowly but thoroughly.