5-Day Demo Crash-Log Challenge - One Repro Folder Per Weekday Before October 2026 Next Fest

Your fest demo crashes. Someone posts a clip in Discord. You ask “what build?” They shrug. You fix a guess. The same crash returns on October 2026 Next Fest day three because repro died in chat.

This Challenges & Sprints piece is a five-weekday challenge: one release-evidence/01-build/crash-repro/ subfolder per day with build hash, steps, logs, and optional replay artifacts—not a promise to fix every bug in five days, a promise to stop losing repro.

Why this matters now (May 2026)

1. Fest demo density — More players, more edge machines, more crash reports.
2. Diligence culture — Publisher packets expect grep-able 01-build/ proof.
3. Maintenance habit — Friday Block 5b keeps folders current; this challenge bootstraps crash discipline.
4. Truth stack — Store truth audit fails if demo stability lies.

Direct answer: Mon–Fri, file one crash repro folder per day (even if “no new crash” = document smoke pass). Sunday export README index.

Why Discord fails as your crash database

Discord threads are great for alerts and terrible for archives:

• Search breaks across channels
• Build numbers scroll away
• Clips expire or compress away steps
• New teammates never scroll history
• Partners cannot grep Discord

Folders in release-evidence/ fix each failure mode. The challenge is behavioral training: post the folder link in Discord, not the screenshot alone.

Who should run this challenge

• Teams with a fest demo branch or playtest branch
• 1–5 person studios without dedicated QA
• Anyone who relied on Discord for crash memory in 2025

Skip if: no runnable demo build this week—fix that first.

Challenge vs bug-fix sprint

Do not merge—run challenge before fix week.

Prerequisites (60 minutes Sunday or Day 0)

• [ ] Fest or playtest branch name frozen
• [ ] release-evidence/01-build/crash-repro/ created
• [ ] player.log / Godot log path documented in README
• [ ] One owner signs folders (engineer default)
• [ ] Discord pin: “crashes go to folder template, not threads-only”
• [ ] Truth audit green or deferred with risk note

Rules

1. One folder per weekday — no doubling Monday crashes on Tuesday’s slot.
2. Build hash in every folder — menu string or version.txt.
3. No fix required same day — repro capture beats hero patching.
4. Honest empty days — no-new-crash-smoke-ok.md counts.
5. No new features during challenge week—stability lane only.

Folder layout

release-evidence/01-build/crash-repro/
  README.md
  2026-05-19-mon/
  2026-05-20-tue/
  2026-05-21-wed/
  2026-05-22-thu/
  2026-05-23-fri/

Each day folder minimum:

# Crash repro — YYYY-MM-DD (Weekday)

- Build: 518
- Reporter: internal / playtest / stream
- Platform: Win64 / Deck / etc.
- Steps:
  1.
  2.
- Expected:
- Actual:
- Logs: player.log path or attach name
- Replay: path or N/A
- Status: open / fixed / wontfix-demo

Day-by-day gates

Time budget per day

Total ~6–10 hours for the week—cheaper than one fest-day fire drill.

What “good” repro looks like

A partner or senior engineer should reproduce in 15 minutes without calling you:

1. Open folder README
2. Install or launch build hash listed
3. Follow numbered steps
4. See failure or attached log matching description
5. Optional: run replay bundle

If step 2 fails, folder is incomplete—not “mostly done.”

Monday — Triage without fixing

Spend 45–60 minutes:

1. Export Discord / Steam discussions into mon-triage-list.md
2. Dedupe by symptom
3. Pick order for Tue–Thu (blockers first)
4. Do not patch yet unless crash prevents any repro

Tuesday — First repro deep dive (model)

Symptom: crash returning to main menu after second floor portal.
Build: 518
Steps: new game → tutorial skip → portal twice → crash
Log: log-tail.txt shows null ref in FloorLoader
Tag: P0
Status: open

Engineer does not fix Tuesday night unless fest is 7 days away—folder still wins.

Wednesday — Platform pass

Run smoke on second-priority platform (often Deck). If new crash, Wednesday folder; if not, smoke doc. Update platform matrix.

Thursday — Third repro or debt paydown

If triage list empty after two repros, Thursday is log hygiene:

• Move stray logs from Downloads into folders
• Delete duplicate Discord threads after linking folder
• Update demo patch notes known issues section

Tuesday–Thursday — Capture discipline

For each repro:

• Clean machine or new user account when possible
• Record effective input paths if control-related
• Attach screenshot or 30s capture filename
• Link Godot threaded load epoch if floor-related

Friday — Index and handoff

README.md table:

| Day | Folder | Build | Status |
|-----|--------|-------|--------|
| Mon | triage only | 518 | — |
| Tue | 2026-05-20-tue | 518 | open |
| … | | | |

Copy summary into operating review Block 1.

Integration with maintenance ritual

After challenge week, Block 5b includes:

• New crash folder since last Friday?
• README index updated?
• Open rows ≤ agreed cap?

When “no crash” days happen

Celebrate with no-new-crash-smoke-ok.md:

- Build 518
- 15-min smoke: menu → demo → quit
- Result: no crash
- Note: does not prove stability forever

Honest empties beat silent skips.

Severity tags (use in every folder)

Tag in folder + README—partners respect honest P3 lists.

Platform matrix (capture once Wednesday)

Deck row ties to input preflight.

Log file map (engine-specific)

Copy tail 200 lines into folder as log-tail.txt—do not paste entire multi-MB logs into git without LFS plan.

Replay hook optional path

If using deterministic replay:

crash-repro/2026-05-20-tue/
  README.md
  input.json
  scene-graph.json
  clip.mp4 (optional)

Even one replay artifact upgrades partner confidence.

Discord → folder workflow

1. User posts clip
2. Moderator replies: template link + “need build #”
3. Engineer creates folder within 24h
4. Thread pinned with folder path
5. Status emoji updated when fixed

Stops infinite “works on my machine” loops.

Coupling with playtest branch

Playtest installs should reference crash template in PLAYTEST.md. Testers who file structured repros are worth ten vague reports.

Coupling with attribution

If crash came from tagged UTM surface, note surface= in folder—helps see if one channel sends broken builds.

Anti-patterns

1. Screenshot-only repro with no build hash
2. “Fixed maybe” without folder update
3. Renaming folders mid-week
4. Fixing crashes without updating index
5. Deleting folders after fix—archive instead

Post-challenge month (June–September)

• Weekly: any new P0 gets folder before patch merges
• Monthly: repeat 5-day challenge if backlog >10
• Pre-fest: cold-open zip includes crash-repro/README.md

Extended FAQ

Streamers crash live?

Log build + timestamp; public clip URL in folder; no obligation to fix before stream ends.

Duplicate crashes same root cause?

One folder; add duplicate-of: tue-folder in notes.

Saves corrupted?

Separate tag save-data; attach save file if small.

WASM / browser demo?

Separate subfolder web/ per WASM trend.

Glossary

Seven-day extension (optional)

Add Sat–Sun buffer for fix sprint after folders exist—challenge week stays capture-only.

Hour-by-hour Tuesday example (solo dev)

09:00 — Reproduce crash once, no fixes.
09:20 — Write steps from memory, reproduce again to verify.
09:35 — Copy log tail.
09:45 — Screenshot or short capture.
09:55 — Commit folder; paste path in Discord.
10:00 — Return to feature work—folder day done.

Partner zip slice (Friday optional)

Zip only:

release-evidence/01-build/crash-repro/

Under 50MB for email—link full repo separately. README first line: build hash.

Regression guard after fixes

When closing P0, add regression line to folder:

## Regression
- Fixed in build 520
- Re-tested steps 2026-05-24: pass

Prevents reopening same ghost in January.

Community moderation script

Thanks for the report! Please include build # from title screen.
We file repros at release-evidence/01-build/crash-repro/ — staff will link your thread when filed.

Reduces duplicate rage threads during fest.

Steam discussion vs private repro

Public review bombs sometimes cite crashes already fixed. Link known issues list and build number in replies; keep full repro in private folder—do not post log tails publicly (spoilers + noise).

Cert and console parallel

Console cert rejection bullets map 1:1 to folders when Quest playbook applies—same discipline, different reviewer vocabulary.

Challenge completion certificate (team morale)

Internal only—post in Discord when Friday index done:

Crash-log challenge ✅ — 5/5 days filed — open P0: N — index: <path>

Cheap morale; reinforces norm.

When to rerun before fest

Stack rationalization note

One-engine teams should document engine log path once in root README—reduces Tuesday confusion.

Key takeaways

Key takeaways

• Five weekdays, five folders (or honest smoke docs)—habit beats heroics.
• Build hash mandatory every time.
• Challenge captures; fix week follows.
• README index on Friday for grep partners.
• Tie into Block 5b maintenance after week one.
• Discord is notification; folder is truth.
• Fest demos fail softly when repro exists early.

Known-issues list for store (Friday output)

Translate P1/P3 folders into player-facing bullet list:

## Known demo issues (build 518)
- Rare crash when re-entering portal twice (fix in progress)
- UI glyph wrong on Deck for map screen (workaround: keyboard)

Feeds truth audit Block 4 honesty.

Jira / tracker bridge

Map folder paths to tickets:

| Folder | Ticket |
|--------|--------|
| 2026-05-20-tue | BUG-142 |

Use playtest triage bridge format—15 minutes Friday.

Contrarian note

Some teams say “we’ll log after fest.” The 2026 counter-argument is visit concentration: one fest hour can generate more crash reports than a month of friends-and-family play—without folders you drown.

Fest week freeze

During Next Fest live days:

• New crashes get folder same day—no exception
• Fixes require new build hash in folder when closed
• No deleting folders mid-fest

Archive policy

When fixed:

crash-repro/archive/2026-05-20-tue-BUG-142-fixed/

Move, do not delete—diligence history matters.

FAQ

How is this different from truth audit challenge?

Truth audit aligns marketing vs build. Crash-log challenge aligns reports vs repro. Run truth first or same week different owners.

Five days enough?

Enough to build habit before fest month. Repeat monthly if needed.

Can we skip Monday?

Monday prevents fixing wrong three bugs—minimum 30 min triage required.

What if we have zero crashes?

Five smoke docs + celebration post—still pass challenge.

Publisher asks for stability proof?

Send crash-repro/README.md + zip slice of 01-build/.

FAQ — search snippets

How do indie teams document demo crashes before Steam Next Fest?

Use one release-evidence/01-build/crash-repro/ subfolder per weekday with build hash, repro steps, log tails, and a Friday README index—five-day challenge habit before October fest traffic.

What should be in a crash repro folder?

Build id, platform, numbered steps, expected vs actual behavior, log excerpt path, severity tag, status, and optional replay artifacts—not screenshots alone.

Should we fix P0 during the challenge week?

If fest is <14 days, yes after folder exists. If fest is >30 days, capture week then fix week—do not skip folders to chase fixes.

How does this connect to AI live-ops?

Prompt promotions mid-week require crash folders before rollback—note prompt-version in repro if dialogue-related.

Public playtest crashes?

Same folder format; redact reporter name if needed.

Console?

Add platform row; attach cert-style logs if applicable.

AI dialogue crashes?

Tag ai-touchpoint per prompt registry.

Build promotion discipline during challenge week

When uploading build 519 mid-week:

1. Create build-519-promotion-note.md in 01-build/
2. Update every open folder header if still reproduces—or close with “fixed in 519”
3. Never leave folders saying 518 if default branch is 519

Crash repro without build discipline recreates publisher diligence confusion.

Soft-lock vs crash taxonomy

Use replay hooks heavily for soft-lock—harder to describe in prose.

Team roles during challenge

Solo: same person, still split hats across days to avoid capture fatigue.

Metrics you are not inventing

Do not publish “crash rate down 40%.” Do track:

• Folders filed / weekdays
• P0 open count Friday
• Cold-open repro success by colleague

Qualitative health only.

Sample triage list (Monday)

# Mon triage — build 518

1. Portal twice → menu crash (P0)
2. Deck map glyph wrong (P1)
3. Audio pop on alt-tab (P2)
4. Wishlist button unrelated (out of scope)
5. Feature request co-op (out of scope)

Items 4–5 go to backlog, not crash folders.

Integration with four-Friday operating adoption

If running operating review adoption, add Block 1 line during challenge week:

Crash-log challenge day 3/5 — open P0: 1

Shows partners ops + engineering alignment.

WASM and dual-SKU caveats

Browser SKU crashes get web/ subfolders—do not mix with PC repro steps without platform header.

Pricing and crash confusion

Regional pricing changes do not belong in crash week—confound signals. If players report “paid but crash,” folder still filed; finance Block 5 separate.

Ninety-minute Day 0 setup

Printable weekday card

Mon: triage list
Tue: repro folder 1
Wed: repro folder 2 or smoke
Thu: repro folder 3
Fri: README index + ops note

Sample Friday operating review line

- Crash-log challenge complete: 5/5 folders; open P0: 1 (floor load); index at 01-build/crash-repro/README.md

Related internal links

• Release-evidence taxonomy
• Playtest triage bridge
• Quest resubmission evidence
• Vertical slice honesty

Sibling challenge comparison

Run crash-log after truth audit when marketing and stability must both be honest before fest.

Legal and privacy note

Redact player names in public-facing known-issues; keep full reporter id in private folder only. GDPR-conscious teams should avoid attaching personal data in log-tail.txt without policy review. When in doubt, store repro steps and build hash only—skip optional attachments until counsel confirms retention rules.

Final upload gate

• [ ] Five weekday entries or smoke docs
• [ ] Friday README index
• [ ] Open P0 count documented
• [ ] Known-issues blurb drafted if needed
• [ ] Block 1 + Block 5b updated
• [ ] Discord template pinned

One completed index beats five heroic all-nighters the week before fest gates open—plan accordingly, then protect the calendar like a ship date.

Close: October 2026 Next Fest rewards demos that fail reproducibly in folders, not dramatically in chat. Run the five-day crash-log challenge while May still has slack—so August diligence and October traffic meet a team that can answer “what build?” in one grep. Archive folders when fixed—deleting repro is how teams fight the same ghost twice. Pair the challenge with Friday evidence maintenance so the habit survives the week after the sprint ends. The squad finishes when the index is grep-ready—not when Discord goes quiet. Start Monday even if your backlog feels infinite; triage is how infinite becomes three lines.