The user who asked this question is not imagining things, and their file is not haunted. What they are describing is a classic symptom of a spreadsheet that is doing far more work than the user realizes, or than Excel is willing to show them. The crash happening twenty minutes after the data finishes loading is not a random glitch. It is likely the result of a background dependency chain that Power Query does not fully release after the initial refresh.
Here is what is probably happening. When Power Query runs those extensive transformations, it builds an in-memory model of the data, and Excel holds onto that model for as long as the workbook is open. Even after the loading bar finishes and the user sees a calm, stable sheet, the application may still be maintaining compressed tables, query metadata, and internal references that were created during the refresh. If the workbook also uses linked tables, named ranges, or formulas that depend on the query output, Excel can trigger a recalculation or a background refresh attempt without any obvious prompt. That hidden activity consumes memory and CPU cycles, and on a machine with limited resources, it eventually exceeds a threshold, and the application crashes.
This matters for every professional who relies on Power Query for heavy lifting. The practical takeaway is that a workbook can appear idle while its engine is still running. Users should treat the post-load period as a risk window, not a signal that the work is done. The first step is to check whether the workbook has any external links, volatile functions, or automatic background refresh settings enabled. Turning off background refresh in Power Query options is a simple fix that prevents Excel from re-executing queries without permission. Another useful habit is to save the workbook, close it entirely, and reopen it after the initial load completes, this forces Excel to release the cached structures and start fresh.
The deeper lesson here is about trust. Traditional spreadsheets give the illusion of simplicity: you load data, you see it, you move on. But as data complexity grows, the tools we use must be transparent about what they are doing in the background. Users should not have to guess whether their file is stable or quietly collapsing. The fix for this user is technical and straightforward, but the broader need is for spreadsheet tools that communicate their state honestly, so that a crash does not feel like a mystery. A tool that leaves you wondering why it failed has already failed its primary job.