Help Index → When something does not behave as expected → Use case 60
Stop additional writes first, then distinguish an unwanted in-memory edit from an uncertain disk save. Protect the known-good backup before choosing a recovery path.
Before you start
- Keep NMS closed. Record the exact error, loaded slot/context and any reported file paths; do not repeatedly retry Save.
- Keep the known-good backup unchanged. Before any synchronization, Pretty Print or structured refresh, copy EACH pending Raw JSON pane to its own external file; syncing one tab can overwrite pending text in all others. Folder backups capture disk, not unsaved memory.
- A parse error can leave the old tree with a green light; [] is ignored, and Pretty Print can overwrite invalid pending text from the old model. Preserve text and verify actual values, not lights. Do not sync or Pretty Print Inventory Raw JSON: its current dictionary-assignment defect is not reliable inventory replacement. Use Organizer for supported moves or Replace Items from Backup for supported replacement.
Steps
- Before syncing, formatting, retrying or reloading, preserve EACH pending Raw JSON pane externally. A parse error can leave a green light beside the old tree; [] is ignored and Pretty Print can erase invalid pending text. Do not sync Inventory Raw JSON as recovery. If inventory replacement failed, capacity may already have expanded before missing Slots caused failure: inspect actual items/capacity, do not save or retry blindly, and deliberately reload the saved baseline if abandoning it. (control reference)
- If no live write was attempted and only app edits are unwanted, preserve other needed work and deliberately reload the intended disk slot. Read the replacement warning before accepting; reloading abandons unsaved app changes, not earlier disk writes. (control reference)
- If a live save failed, read whether recovery also failed. The writer attempts to restore the immediately pre-write data/meta pair after certain caught failures, but that is not a full backup or a guarantee against every interruption. Treat a recovery-failed warning as uncertain disk state. (control reference)
- For full-folder recovery, open Full Live Save Restore and read its Manual Instructions. Check the displayed runtime live directory and backup root, then identify your specific known-good backup. Copy Text only copies instructions; there is no restore execution button. (control reference)
- Before performing any manual replacement outside the app, preserve the current failed folder separately and verify both source and destination. The displayed destructive example removes live-directory contents. Stop if paths or backup contents are uncertain; do not run an example blindly. (control reference)
- After manual restore, reload the intended slot and inspect its context before editing. At the unsaved-close warning, Cancel keeps work open; Close discards it without saving. (control reference)
Expected result
- Recovery follows the identified failure scope while keeping the known-good backup separate from both failed disk files and unsaved editor text.
What changes, and when to save
Manual restore instructions neither restore nor save anything. Do not File → Save stale in-memory data over a restored folder; reload first. A new backup does not recover an earlier version automatically.
Check the result
- Confirm the reloaded slot/context and expected pre-edit values. Do not declare recovery from a copied command, closed dialog or success count alone.
Important details and recovery
- Never reuse the known-good backup as the destination for preserving a failed folder.
- The app has no universal undo-all. If recovery remains uncertain, keep NMS closed and retain all copies for diagnosis.
- When retaining understood non-inventory text edits after a trustworthy baseline is restored, reapply from external copies one pane at a time, sync and verify actual values before moving to another pane. Never just sync each pending tab; the first sync can overwrite the others.