Audit

I dislike unexplained observations.

Not because they are frightening.

Because they are inefficient.

Every unexplained observation generates unnecessary possibilities.

The reappearing cross-reference left me with only one sensible course of action.

Assume the error belonged to me.

I began with the obvious.

I compared every available revision of the index.

Nothing.

I checked archived exports.

Nothing.

I rebuilt the directory structure from an earlier backup.

The reference remained.

At that point I stopped trusting my own memory.

Fortunately, computers are less imaginative.

I began comparing timestamps instead.

Every directory. Every modification date. Every recorded revision.

Nothing appeared inconsistent.

I wrote a script to search for every filename, every fragment of the title, every identifier and every variation I could reasonably generate.

No result.

The destination existed only as a destination.

Never as a file.

I disconnected every device from the network. Worked locally. Compared the results.

No change.

Several times I convinced myself I had finally found the explanation.

Each time it dissolved under closer inspection.

The goal of an investigation is not to defend your favourite explanation.

It is to eliminate explanations until only those supported by the evidence remain.

At the end of the audit I was able to conclude exactly three things.

First. The reference behaved consistently.

Second. The referenced file did not exist anywhere I could locate.

Third. I could no longer reproduce the circumstances under which the reference first appeared.

Everything beyond those three statements would have been speculation.

So I stopped.

I recorded the inconsistency.

Marked it unresolved.

Continued working.

That decision turned out to matter far more than I understood at the time.