The problem

“Online” is too coarse to explain a two-second failure.

GapTrace was motivated by intermittent mobile-tether and game-connectivity problems where ordinary ping tests could look healthy while DNS, TCP or HTTP reachability failed briefly.

The probes run concurrently and remain deliberately unprivileged: no raw-ICMP requirement, service installation or administrator-only capture layer.

Failure model

Record symptoms precisely without pretending to know the carrier root cause.

01

DNS

Separates name-resolution failure from broader connectivity loss.

02

TCP

Tests practical socket reachability independently of DNS success.

03

HTTP

Adds an application-layer reachability signal instead of relying on one network primitive.

04

Analysis

Extracts outage events, recovery durations, availability and latency distributions from recorded evidence.

Measured release

Background diagnostics should not become the workload.

The 1.0 release was validated locally with a sustained monitor run and a packaged executable smoke test.

peak working set16.2 MB
CPU time · validation0.016 s
release EXE6812.5 KiB
release ZIP2.8 MiB
frozen commitb77911d
release stateVALIDATED 1.0.0

Focused utility

Capture the transient failure first. Diagnose from evidence second.

GapTrace is useful independently, and its structured evidence can later become an optional low-overhead network source for CrashScope.