Continuation prototype profiling showed scalability issues with ChunkPool implementation.
Note: Resources are semantically thread local. Continuation usage was very specific to internal vm processing for freezing/thawing and has since been removed due to the cost. The profiles highlighted a high cost of freeing chunks. Need to create some microbenchmarks to determine if this is a problem with heuristics of freeing chunks or more fundamental issue with a shared global lock across ChunkPools. Sample profile attached.