libvips / pyvips latency with concurrent `write_to_memory` calls

Question

I\'m trying to generate small image tiles from a larger multi-channel tiff image concurrently and running into large delays on the first few shrink.write_to_memory() calls (