Radix Week in Review: Nine Accounts and Four Fees of 100%

Week in brief

• hyperscale-rs merged its simulation-testing fixes and an experimental network layer that connects nodes through home routers, in 121 commits.
• A Radix Guild escrow contract, ported to the hyperscale VM, ran across four shards and found two bugs.
• Nine accounts that left the offline Weft validator on 7 and 8 October unstaked 326m $XRD more from 13 validators; staked $XRD fell 7%.
• Four validators queued fee rises to 100% and one to 75%; one account signed four of the five requests.
• $XRD traded near 0.00015 US dollars on KuCoin, about half its price a week earlier.
• OTER is finishing its contracts for mainnet, and candidacy for the Permanent RAC is open to 30 October.

Most of the code written for Radix this week went into hyperscale-rs, the Rust implementation of a sharded Radix network, and most of it was about what happens when things go wrong: dropped messages, cut links between shards, peers behind routers that refuse incoming connections. A community developer used the same tools to run a real Radix contract across four shards and found two bugs on the first attempt.

On the live network, stake moved out. The same accounts that unstaked 109m $XRD from the stopped Weft Finance validator the week before unstaked three times as much from 13 healthy ones, and five validators asked to keep most or all of the rewards their delegators earn.

Hyperscale: testing for the rare failure

flightofthefox, who writes hyperscale-rs, merged two long-running branches. PR #171, merged on 16 October, collects fixes found by deterministic simulation testing, which replays thousands of slightly different faulty networks from a seed so a failure can be reproduced exactly. PR #172, merged on 17 October, adds an experimental network layer built on iroh, a Rust library that connects two machines directly even when both sit behind home routers, and falls back to a relay when it cannot. The node now chooses between iroh and its existing libp2p layer at build time, and a container testbed seats a committee of validators across different router types.

A Guild contract on four shards

bigdev_xrd ported the Radix Guild's escrow contract to the hyperscale VM and ran it on hyperscale-rs across four shards: 21 steps passed on five seeds, including one transaction spanning three shards that rolled back cleanly when a later step failed. The run found an account that could not move its second NFT of the same kind. A fault test found the other: if certificate traffic between shards stays cut for about 20 minutes, a cross-shard transaction can stay locked for good, with its funds held but not lost. A commit on 16 October stops a certificate from deciding anything across shards, the change flightofthefox described as the fix.

Who keeps the history

Seböööl measured the Babylon ledger at 730 GiB, of which 16.25 GiB is the state a new node needs; 89.6% of transactions touch at least two components. flightofthefox separated two jobs: archival, keeping old transactions and receipts, which Babylon gives to every validator although consensus does not need them; and indexing, arranging them so a wallet can ask for its last 50 transactions, which is the Gateway's work. On hyperscale, archival would be an opt-in node role, and his sketch has archive nodes post a bond, with the rewards that bond would have earned staked redirected to the nodes that answer their queries. Asked when the network would reach an operational test, he answered:

Nine accounts, 326m $XRD

Unstaking on Radix is two steps. An unstake burns the holder's liquid stake units and issues a claim NFT; the $XRD can be claimed 2,016 epochs later, about seven days. On 7 and 8 October eleven accounts unstaked 131m $XRD from the Weft validator, whose node had stopped voting. Between 13:56 UTC on 16 October and 14:11 UTC on 18 October, nine of those eleven unstaked 326m $XRD from 13 other validators, in 27 transactions, each splitting its exit across one to four validators in round amounts. All unstaking on the network over the week came to 377m $XRD, so these nine made up 86% of it.

Where it came from and where it went

Reddicks' validator lost 113m $XRD, half its stake, and fell from third to seventh; Avaunt Staking lost 47m and GenkiPool 40m. $XRD the nine had claimed from Weft was already moving. One of them claimed 8.7m on 16 October and the next day deposited 6m into an intermediate account, where two more of the nine had put 7.8m; that account passed each deposit on within three minutes, less 0.5 $XRD, to a single collecting account (one of them). Exchange deposit addresses forward funds this way.

The ledger does not say who controls the nine. Holders in the main Radix group attribute recent selling to wallets they link to RDX Works, starting from a September 2024 transfer that split 1.17B $XRD into 16 accounts (Jelthebest). Read on 18 October, those 16 accounts hold no $XRD, and none of them is among the nine. That evening Gary, a holder who had offered to buy out the remaining sell pressure, wrote that a few hundred million $XRD had already been unstaked.

Fee requests to 100%

A validator's fee is the share of its delegators' staking rewards the owner keeps. A rise takes effect 4,032 epochs after it is filed, about 14 days, so delegators can leave first, and a validator holds one pending change at a time: a new request replaces the old one. Five requests filed on 17 and 18 October use that notice.

Sirius and Polaris are run by Cadwynbloc, whose site both name. One account signed the requests for Sirius and Polaris (17 October), Cadwynbloc (18 October) and Radstakes (18 October), the same account behind the September rises and the 50% requests of 9 October, which the new ones replace. RadixTalk's came from a different account. At 100% the delegators of a validator earn nothing on their stake.

Markets: one token, several prices

$XRD fell from about 0.0003 US dollars on 11 October to 0.00015 on KuCoin late on 18 October, by CoinGecko's figures. Gate, where $XRD deposits have been closed since at least 11 October, quoted 0.0005 at the same time, and CoinGecko flags that quote as an anomaly: with deposits shut, nobody can buy elsewhere and sell on Gate to close the gap. Holders noticed wallets showing different prices depending on which venue they draw from.

In the main group, 合lex keeps a dashboard of the accounts holders suspect of selling and raised its estimate of what may still come by about 600m $XRD on 18 October. Gary said he had been trying to arrange a buy-out of the remaining $XRD and would decide this week whether to keep making offers. Ociswap volume fell 47% over the week.

Built this week

OTER is finishing the roughly 17 contracts it plans to put on mainnet. Daffy counted 317 deployments of its oracle blueprint on Stokenet in eight days, 105 in one; the team explained that its end-to-end suite deploys the whole stack on each of seven test lanes and will now publish each package once per build.

bigdev_xrd open-sourced scrypto-audit-kit, which checks Scrypto blueprints with static analysis and an LLM checklist and can record an attestation on the ledger. A fuzz harness he wrote for radixdlt-scrypto found two low-severity bugs in developer tooling, and he asked where to report them: the repository has no security policy and the DAO's disclosure channel is not live. GenkiPool added pool, locker and deposit-rule tools to its browser console, and RSwap put part of its liquidity back on 16 October.

This week on the wiki

How to Stake and Unstake $XRD, new on 15 October, sets out the two delays both of this week’s ledger stories run on: 2,016 epochs to claim unstaked $XRD and 4,032 epochs of notice before a fee rise. A delegator on one of the validators above has until about 31 October to move. How to Start a DAO on Radix also arrived, comparing CrumbsUp, Caper and Muan.

Pages edited in the seven days to 2026-10-18:

The week on the ledger

Read live from the Radix Gateway at epoch 350,682, state version 564,201,503, on 2026-10-18.

The ledger

One claim settled, and the record stands at 12 hit, 4 miss and 3 open (all claims).

• Sirius and Polaris to 50% at epoch 352,012: miss. The account that filed the request replaced it on 17 October with one for 100%, read at epoch 350,682 through the Gateway and radixscan; 352,012 will pass with no change.
• New: all five fee rises in the table above take effect as queued, settled by 8 November on each validator’s effective fee at or after its epoch.

Concentration watch

8 validators hold a third of all staked $XRD, the same count as last week; a group holding more than a third can stall the network. It is not the same eight: Avaunt Staking dropped out after losing 47m $XRD, and RadixStake took its place.

What to watch

• About 23 to 25 October: the 326m $XRD unstaked on 16 to 18 October becomes claimable; next week’s reading shows whether it is claimed and where it goes.
• About 31 October and 1 November: the five fee rises reach their epochs, settled by 8 November.
• 30 October: candidacy for the Permanent RAC closes; the Transition RAC wants at least 12 candidates.

Corrections

Nothing this week.

Next week the 326m $XRD becomes claimable, and the fee rises are a week from landing.

Radix Week in Review series, Issue #17: ← Previous: Issue #16, Oct 5–11, 2026 · All recaps · Subscribe

Previously: The stuck hUSDC has a refund list, 131m XRD left an offline validator, and two validators queued fee rises to 50%.