Have you tried aggregated capture files? https://bloomberg.github.io/memray/run.html#aggregated-captu...

With that option the files are much much smaller and much easier to analyse