Bottom line, it looks to be more risky than I thought, and we don't have time to investigate before 5.9rc, so I've changed it to

In fact, python does eventually collect garbage on its own, but it takes like 15 requests. So full-garbage-collection-rate of 100 (let alone 1000) doesn't change anything, since the process will either collect on its own, or it'll grow in size so much that it gets a visit from OOM killer.

@av6 thanks for the detailed perspective. I will make testing along the lines you suggest tomorrow (it's late here, now), but this here is worrying:

Depends on the definition of NodePrefixRef. I don't think there would be any extra allocation if you define it like this:

@martinvonz yeah, probably wrote the doc-comment too fast

As the new commit description explains, I've done all I could to make the change of hash format easier

These are done, thanks for the remarks.