Context propagation is the thread that stitches spans from many services into one trace; break it and the trace fragments.
Trace context propagation is the passing of a request's trace and span identifiers across service boundaries, typically in W3C traceparent headers, so that spans emitted by different services all join the same trace.
When a service makes a downstream call, it injects the current trace context, the trace ID and the active span ID, into the request, usually via the W3C traceparent header. The receiving service extracts that context and makes its spans children of the caller's span. The standard header format is what lets tracing work across languages and vendors.
If any hop fails to forward the context, the trace splits: the downstream work shows up as a separate, orphaned trace, and you lose the end-to-end picture exactly where you need it. This is the most common reason traces look incomplete, and why context propagation matters as much as the instrumentation itself.
Ops Singularity relies on well-propagated context to correlate a symptom in one service with its cause in another; where context flows correctly, Sentinel AI can trace an incident across the whole request path.
A W3C-standardised HTTP header that carries the trace ID and parent span ID between services, enabling interoperable context propagation.
Usually because trace context is not propagated at some hop, so downstream spans start a new trace instead of joining the original one.
Ops Singularity turns open telemetry into autonomous, governed resolution. See it on your own stack.