The heap dump v2 format was designed to support incremental heap dumps [allows nodes, strings to  be shared across heap dumps]. This reduces the size of traces that contain multiple heap dumps. Currently, this is supported by chrome://tracing UI and symbolize_traces.py, but not supported by the OOP HP heap dump exporter, nor the memlog server processing.

This sounds like a good optimization for traces that have multiple heap dumps, but will add non-trivial complexity to the OOP HP heap dump exporter. If we do choose to add support for this feature, we should make sure that there's test coverage in both the exporter, and all the tools that process traces..