When I adapted the TweetNaCl library and included it in the micropython runtime, it took over a second to make or validate an ed25519 signature. For reference, ordinary PyNaCl takes a few dozen microseconds for either (29 μs for sign and 58 μs for verify on my machine; just now benchmarked it). A millisecond is probably the most latency that can be tolerated without making the router logic into its own denial-of-service.
I can currently see a few possible paths forward:
Already started the fork of libsodium here but have not gotten very far with it.
Pay now to fund the work behind this issue.
Get updates on progress being made.
Maintainer is rewarded once the issue is completed.
You're funding impactful open source efforts
You want to contribute to this effort
You want to get funding like this too