Office of the Cosmic Ray Ombudsman Receiving claims · ground station, 11 m
Adjudication

Something flipped a bit. It probably wasn't the sky.

Every claim that a defect was caused by a cosmic ray is assessed here against the particle flux where the machine was actually standing. The mechanism is real and well measured. Your claim, in all likelihood, is neither.

1,204Claims heard
1Claims upheld
23.26Years since the last one
×2Flux per 1,500 m of altitude
Claim assessment

File a claim against the sky

A cosmic ray striking the upper atmosphere produces a shower of secondary neutrons, and a neutron crossing unprotected memory can deposit enough charge to flip a single bit. Error-correcting memory exists for that reason alone. Enter the conditions and the office will rule.

No claim filed. The office is receiving.
Neutron flux where the machine stands Relative to sea level · log scale

Bars are logarithmic. At cruising altitude the flux is roughly 134 times what it is at sea level, which is why the claim is easier to believe from a plane.

Case record

The one claim this office has upheld

+4,096 Surplus votes recorded
Schaerbeek, Belgium · 18 May 2003

A candidate received more votes than the commune had voters

The surplus was exactly 212: the thirteenth bit, set where it should have been clear. It was detected only because the arithmetic became impossible. The office keeps the file permanently on display, on the grounds that it is the only occasion on which the excuse was true, and because a defect that announces itself in a clean power of two is considerably rarer than the excuse.

Standing guidance

Why your claim will be dismissed

01

It reproduces

A particle passes once and does not wait for your test suite. If the fault returns on demand it was written by someone, reviewed by someone else, and approved by both.

02

Your memory would have said so

Error-correcting memory catches a single-bit upset and records it. A machine with ECC and an empty log is a machine reporting that nothing arrived.

03

The measurements disagree with you

Large-scale study of memory in service found errors far more common than predicted, and predominantly hard rather than transient: the same cells failing repeatedly. That is a component wearing out, and unlike the sky it can be inspected.