Distributed traces are crucial for observing distributed systems and microservices, but storage limitations necessitate sampling methods like head-based and tail-based sampling, which have their drawbacks. Head-based sampling involves random decisions at the root span, leading to potential under-sampling of rare traces, while tail-based sampling offers more intelligent decisions at the cost of increased memory and network overhead. To address these limitations, Dynatrace is exploring partial trace sampling, which adjusts sampling rates based on the frequency of trace parts, ensuring that less frequent trace sections are sampled more often, providing a more balanced view. Although this approach results in partially sampled traces, the data is still valuable for specific queries. The concept is part of ongoing research and is influencing the development of a new OpenTelemetry sampling specification.