We were hoping to have the CLN release ready for you today, but the deadline gods were against us and we still have an unacceptable number of errors with non-contactable nodes when we scale up

A word from @Nic_Dorman about our approach.
We have been reviewing our approach to node support on the network and need to adjust how we support connections to ensure reliability across uploads and downloads.
To do this we are implementing some changes to remove Non-Contactable nodes (NC, a change in nomenclature from āprivate nodesā which was ambigouous) from future dial attempts by other nodes. This will initially be via a dial back of 180s after which a node that fails to respond will be deemed NC. This will have an impact on nodes that are run from poor quality connections and those that are behind some consumer networking setups (such as those without port forwarding)
We are looking at changing how we undertake some networking operations as well, likely looking at forking libraries we have built on top of to cater to our specific needs.
There will be other changes coming to verify and validate uploads to ensure reliability with the north star of development being data permanence.
Comms and community
On comms, take a look at @jimcollinsonās latest video on just whatās wrong with the online oligarchy.
On community efforts, thereās been some noble work by @safemedia to integrate dweb into the autonomi browser extension. Still some wrinkles to iron out but itās great to see these sorts of efforts.
Another AntTP update from @traktion brought the ability to build as a library or an executable, to allow other apps to integrate AntTP into them as a native library, support for Swagger UI support, plus API support for chunks, private scratchpads and public scratchpads (to compliment pointers, registers and public archives).
And three updates this week from @happybeing with the latest dweb (v0.9.1 bringing publishing and viewing to Windows, MacOS and Linux.
Many thanks to @riddim for the PR suggesting a retry strategy for uploads. This has now been implemented by the team.
General progress
@chriso has been working on the next release candidate, with @shu tweaking the monitoring accordingly. They are also leading the investigations into the scale-up errors.
@anselme has been working on refactoring the node networking stack and simplifying the CLN. He moved the common network constants used in client and node to the sn_protocol
crate to avoid duplication. His next step is to further divide the driver module (and SwarmDriver) into sub-modules with very delimited responsibilities so we can then deal with some complexity issues we have at present.
@bzee made changes to get Project Dave working on the main network again, bringing it inline with our latest client code.
Ermine continued working on replacing RPC with metrics services. He made progress on bringing in a rpc_client
supported service in a way thatās backwards-compatible.
Lajos finished the paymaster PoC and tested a bunch of edge cases on Sepolia. He demonstrated the ability to send ANT data payments using this system by relying only on Ant - no Eth required for gas!
@qi_ma raised a PR to address some hourly CPU usage spikes weāve been seeing introducing a use a random activity timeout to ensure that the nodes do not sync on work, and another two [1] [2] to tackle a high error rate in file downloads caused by failures in the verification process and to improve uploads. This seems effective and errors are vastly reduced in tests. Qi also made some changes to the CLI following a suggestion by @riddim with a new option for retrying PUTs in the event of initial failure. This means users no longer have to go through the whole process of encryption, quoting and uploading chunks again if an upload fails.
@roland pushed a large PR to make the NodeRegistry concurrently accessible. Previously data was passed as &mut
and this limited us to performing / monitoring just one service at any time, a bottleneck for the reachability check work.
And @vphongph has been working on continuous integration (CI), removing unnecessary processes to speed performance and make CI more reliable.