…emory (RaoFoundation#2724)

Archive nodes serving sustained RPC load show unbounded anonymous-memory growth
(~3-4 GB/h) that fits the glibc-malloc arena-fragmentation profile across the
node's hundreds of threads; the release binary, unlike polkadot, is not linked
against jemalloc. Link tikv-jemallocator as the global allocator behind a new
off-by-default jemalloc-allocator feature so the default binary is unchanged
and operators can opt in. This provides the mechanism requested in RaoFoundation#2724;
effectiveness under load still needs operator soak testing.

… test

The off-by-default `jemalloc-allocator` feature is never compiled by the
default build, so a `tikv-jemallocator` bump, a `#[global_allocator]` typo,
or a `cfg` mistake would break the opt-in allocator path with nothing
failing.

Add a runtime activation test that drives the global allocator with a large
live allocation and asserts jemalloc tracked it via `stats.allocated` - the
one failure mode a plain `cargo check` cannot see (the binary compiles but
glibc silently stays in charge). `tikv-jemalloc-ctl` (the canonical
companion crate, same 0.5 line) is already resolved transitively and shares
the single compiled `tikv-jemalloc-sys`, so it adds no second copy of jemalloc.

Gate the feature explicitly in check-rust.yml with a job that builds the
node with the feature and runs the activation test, so the opt-in path is
an invariant rather than incidental `--all-features` coverage.

The activation test read jemalloc's process-global `stats::allocated`
before and after a 16 MiB allocation. `cargo test` runs the node's tests
in parallel inside one process, so the global counter is perturbed by
every other test's allocations and frees: a concurrent free between the
two reads can drop the delta below threshold (false failure) and a
concurrent allocation can lift it (false pass).

Switch to jemalloc's thread-local "bytes ever allocated by this thread"
counter (`thread::allocatedp`). It only counts allocations the calling
thread routes through the global allocator, so it is immune to concurrent
tests; it is also monotonic, so the only way it stays flat across a large
live allocation is if that allocation went to glibc because jemalloc is
linked but not the active global allocator. Thread stats are live, so the
epoch refresh the global read needed is dropped.

…ration test

The doc comment typo fix propagates into runtime metadata, so the generated
Python calls.py must be regenerated to keep the metadata drift gate green.
Also silence clippy::deprecated on the LastTxBlockDelegateTake migration test,
which intentionally exercises the deprecated storage aliases.

Co-authored-by: Cursor <cursoragent@cursor.com>

…ials

With ~120 parallel e2e jobs pulling the same localnet image, GHCR
occasionally returns a spurious permission_denied. Retry with backoff
instead of failing the shard on the first attempt.

Co-authored-by: Cursor <cursoragent@cursor.com>

… docs

- uv publish --check-url: the rc publish uploaded the SDK wheel and then
  failed on bittensor-core (missing trusted publisher); a plain re-run
  would 400 on the duplicate wheel. --check-url skips files already on
  the index so re-runs after partial uploads just resume.

- copilot-instructions.md still described the pre-release-train branch
  strategy (PRs target devnet); the auditor cited it to block a routine
  PR against main. Align it with common.md: PRs target main, the network
  branches are ruleset-locked CI mirrors.

Co-authored-by: Cursor <cursoragent@cursor.com>

…n-hardening

ci: idempotent PyPI publish re-runs; fix stale branch-strategy review docs

…sion

- Bump runtime spec_version 429 -> 430 for the next mainnet upgrade.
- Add the devnet network endpoint to the Python SDK and network docs.
- Read the Python SDK __version__ from installed package metadata so wheels
  report the version they were published as.

Co-authored-by: Cursor <cursoragent@cursor.com>

chore: rao-release — spec_version 430, devnet endpoint, SDK version from metadata, e2e de-flake

Co-authored-by: Cursor <cursoragent@cursor.com>

…ce-check

ci: print CI key address and precheck balance in mainnet proposal

About-style page covering the pending spec 430 upgrade: conviction
subnet ownership, price-driven emissions, bittensor v11, Ledger and
extension signing, the sign-and-verify upgrade flow, and required
actions — all linking into the docs. Includes an OG share thumbnail
and a production toggle on the Deploy Docs workflow so the website
can ship without waiting for the chain upgrade.

Co-authored-by: Cursor <cursoragent@cursor.com>

Co-authored-by: Cursor <cursoragent@cursor.com>

/releases lists all network releases newest-first (data-driven, one
array entry per release) and links to each release page. The docs
sidebar gains a Releases link via a meta.json external link entry.

Co-authored-by: Cursor <cursoragent@cursor.com>

Byline and essay voice matching the About page; the upgrade is
described as shipped rather than pending.

Co-authored-by: Cursor <cursoragent@cursor.com>

Remove first-person voice and hyperbole in favor of a formal,
explanatory register with documentation links throughout.

Co-authored-by: Cursor <cursoragent@cursor.com>

Co-authored-by: Cursor <cursoragent@cursor.com>

Upgrade commands, a v11 plan/execute example, and btcli usage taken
from the migration guide and quickstart.

Co-authored-by: Cursor <cursoragent@cursor.com>

Inline SVG showing total conviction crossing the 10% threshold after
the one-year age gate, with the contestable region highlighted.

Co-authored-by: Cursor <cursoragent@cursor.com>

Co-authored-by: Cursor <cursoragent@cursor.com>

Bar comparison of three same-price subnets: shares shrink with age
under the old root-proportion term, and are equal under the new
price-only split.

Co-authored-by: Cursor <cursoragent@cursor.com>

Co-authored-by: Cursor <cursoragent@cursor.com>

Unit-safe money, policy guardrails, typed errors, signed requests,
and timelock encryption, each linking to its documentation.

Co-authored-by: Cursor <cursoragent@cursor.com>

Proxies, multisig, atomic batches, MEV-shielded submission, key
rotation, and address-book resolution; note that the CLI is generated
from the intent catalog.

Co-authored-by: Cursor <cursoragent@cursor.com>

Runtime-discoverable operation catalogs, execute-by-name, previews,
policy bounds, and machine-readable docs endpoints.

Co-authored-by: Cursor <cursoragent@cursor.com>

Cap code blocks at the viewport width so their pre-formatted lines
scroll instead of forcing the whole page to overflow horizontally, and
bump SVG chart label sizes at phone widths where the scaled-down 10px
labels became unreadable.

Co-authored-by: Cursor <cursoragent@cursor.com>

Stake intents gain a 5% rate tolerance by default with a
slippage_protection escape hatch, SlippageTooHigh gets a targeted
remediation message, and the fake substrate harness plus unit tests
cover the new behavior.

Co-authored-by: Cursor <cursoragent@cursor.com>

The pending SDK changes ship with this release, so the blog, route,
sitemap, releases index, and OG image move from v430 to v431. Drop the
redundant opening sentence.

Co-authored-by: Cursor <cursoragent@cursor.com>

Co-authored-by: Cursor <cursoragent@cursor.com>

Co-authored-by: Cursor <cursoragent@cursor.com>

Co-authored-by: Cursor <cursoragent@cursor.com>

Co-authored-by: Cursor <cursoragent@cursor.com>

The stake tx pages and intents catalog pick up the new
slippage_protection and rate_tolerance parameters, fixing the docs
drift gate.

Co-authored-by: Cursor <cursoragent@cursor.com>

…-page

Add v430 release announcement page at /releases/v430-upgrade

Co-authored-by: Cursor <cursoragent@cursor.com>