Paseo is the project that maintains the Polkadot Testnet. As of today the testnet hosts:
- the Polkadot Products Devnet;
- Parity’s staging environment;
- community parachains.
In Q3 Paseo restarted the Testnet’s relay chain from block 0 on a smaller footprint, as planned in the Q3/Q4 2026 funding proposal (referendum #1919). Asset Hub and People kept their state, Bulletin joined as a system chain.
This post summarises the quarter:
- what runs on the Testnet today, and who runs it;
- how the Testnet did against the SLOs agreed in the proposal;
- how validators performed;
- how runtimes are now released;
- what the Polkadot Testnet dashboard monitors;
- how the quarter was paid for.
At a glance
| 1 July (legacy relay) | 30 September (new relay) | |
|---|---|---|
| Validators in the active set | 152 | 58 |
| Cores | 56 | 19, all assigned |
| System chains | 5 | 3: Asset Hub, People, Bulletin |
| Infrastructure providers paid by the bounty | 27 | 22 |
| Nodes run by providers | 172 | 76 (58 validators, 6 collators, 12 RPC nodes) |
| Quarterly operating budget | 196,250 USDC | 100,400 USDC (−48.8%) |
The reset
The legacy relay was shrunk in steps while the new one was brought up, and teams moved across at a single announced cutover on 6 July. Collectives, Coretime and Bridge Hub were retired. Cores are now assigned directly on the relay with coretime.assign_core, so an allocation stays in place until it is changed, with no sale cycle to renew.
What runs on the Polkadot Testnet
Every core on the relay is currently in use, shared between:
- the Testnet’s system chains: Asset Hub, People and Bulletin;
- Parity’s staging environment: the Next chains, ahead-of-release builds of Asset Hub, People and Bulletin;
- community parachains: 6 different chains as of today.
The Polkadot Products Devnet runs on the Testnet’s system chains.
This devnet is where any developer can find a sandbox to develop products with the new tech stack arriving at Polkadot. As of Asset Hub block #15,062,015 (29 October, 13:04 UTC), the community project Devnet Apps counts:
| .dot names registered | Apps deployed | Publishers |
|---|---|---|
| 476 | 388 | 141 |
We were also able to account for 988 user registrations on the devnet at the time of writing.
SLO results
The proposal sets service-level objectives for the relay chain, the system chains (Asset Hub 1000, People 1004, Bulletin 1010), validators, runtime upgrades, incident response and observability. Results for Q3:
| SLO | Target | Q3 result |
|---|---|---|
| Relay chain availability | 99.5% monthly | Met every month. Since the cutover on 6 July the relay never went a minute without a block; its longest gap was 24 seconds. |
| System chain availability | 99.5% monthly | Every system chain above 99.5% over the quarter, counting stalls of 5 minutes or more: Asset Hub 99.59%, People 99.75%, Bulletin 99.99%. |
| Validator participation | ≥95% of the active set online | ≥95% in every era from 7 September to the end of the quarter. |
| Runtime upgrades (fallback) | Fellowship release deployed within 7 business days (95%) / 10 (99%); no Friday upgrades | Runtimes now ship on Paseo’s own schedule, ahead of Fellowship releases (see Runtimes). No upgrade was enacted on a Friday. |
| Incident response | P0 acknowledged ≤30 min, mitigation plan ≤60 min | Since 11 September a chain stall pages the core team on Telegram at the next 5-minute liveness check. |
| Observability coverage | 100% of relay and system chains | Met. Liveness and RPC reachability checks for the relay and all three system chains, and per-validator authoring and backing observability, all on the Polkadot Testnet dashboard. |
Validator performance
Validator performance improved compared to previous quarters. Over Q3, A+ was the most common grade in every era we read, and A or A+ accounted for almost all grades:
- 77.4% of the 15,292 validator grades issued from 23 July to 30 September were A+, and 93.8% were A or A+.
- In 20 eras every validator in the active set was graded A or A+.
- On 29 September (era 3624), 49 of 56 validators were graded A+ and the other 7 were graded A.
- From 7 September (era 3534) to the end of the quarter, at least 95% of the active set participated in every era, and 94.7% of grades were A or A+.
Across Q3, validators cast 97.5% of the 41.8 million backing votes they were assigned, and confirmed parachain data availability in 98.3% of bitfields.
The participation SLI counts a validator as participating when it misses at most 20% of its backing votes in an era.
Grades are collected from ONE-T by Turboflakes (paseo-onet-api.turboflakes.io), which scores each validator per era from its missed-vote ratio and bitfield availability. The dashboard’s participation page, polkadot-testnet.r0gue.io/participation, tracks the SLI hourly and alerts the team when it drops below 95%.
Runtimes: how Paseo releases now
Until Q2, Paseo’s runtime work was to follow each Polkadot Fellowship release. From Q3, the Testnet’s runtimes are developed, versioned and released from paseo-network/runtimes on Paseo’s own schedule to accommodate for a faster pace, so new features reach the Testnet without waiting for a Fellowship release. The 7-business-day SLO stays in place as a fallback.
The release process:
- A regular cadence. Paseo keeps a regular cadence unless a blocker stops it. Each week the Asset Hub and People runtimes are synced with the latest features and versions of Individuality, and the Bulletin runtime with polkadot-bulletin-chain. Recent examples: #452 and #453.
- Fellowship as the production reference. From the Fellowship runtimes Paseo takes features that differ from Individuality and, above all, runtime configuration, so the Testnet mirrors Polkadot production as closely as possible.
- On-demand releases. Outside the cadence, a runtime is released on demand when it needs an urgent change, or when its configuration gets in the way of the community building its products.
How an upgrade is enacted:
- Build. Every runtime is built with srtool, so anyone can reproduce the wasm and check it against the published digest.
- Static gates. The wasm hash must match the srtool digest, the spec version must move forward, and no other upgrade may be pending. Host functions and runtime-API versions are diffed against the live chain, and the claim-queue offset that stock collators will derive is computed and checked.
- Rehearsal. The upgrade is rehearsed on a fork of the live Testnet (relay, Asset Hub, People and Bulletin), using the same collator binary providers run in production.
- Enactment. The upgrade is enacted, then backing, inclusion and finality are watched until the chain produces normally on the new code.
Weights have been regenerated on a dedicated benchmark machine since 10 September.
Releases and upgrades in Q3:
- 12 releases (v2.3.1 to v2.5.3) and 42 merged pull requests.
- 14 runtime upgrades enacted: 6 on Asset Hub, 7 on People and 1 on Bulletin.
- The system chains moved to polkadot-sdk stable2606 and Fellowship v2.4.0.
- Q3 releases carried functionality the Fellowship runtimes did not have at the time, now present in Polkadot:
- the Individuality (Proof of Personhood) subsystem on Asset Hub and People;
- Bulletin as a system chain;
- the dotNS gateway;
- paying contract fees in PGAS.
Monitoring:
The proposal commits to reachability checks and validator observability for every critical component. During Q3 the Polkadot Testnet dashboard became the network’s monitoring and operations console. It now runs these jobs:
| What | How often | Alerts | Since |
|---|---|---|---|
| Chain liveness for the relay, Asset Hub, People and Bulletin: asks every RPC endpoint for the latest finalized block and flags a chain that stops advancing | every 5 min | Pages the core team on Telegram and posts to Matrix | 13 Jul (Telegram 11 Sep) |
| RPC health of every public endpoint: up/down and whether it serves archive data | every 5 min | Matrix alert that mentions the endpoint’s operator after 15 minutes down | 16 Jul (alerts 27 Jul) |
| Validator participation SLI against the 95% target, with each validator’s ONE-T grade | hourly | Matrix alert below 95%, critical below 90% | 27 Jul |
| Backing health per validator group and per validator | hourly | Shown on the dashboard | 4 Aug |
| Backing report naming validators that miss more than 20% of their votes, with their operators | Mon and Thu | Matrix, operators mentioned | 14 Sep |
| Relay view of the system chains: parachain heads and backed candidates for Asset Hub, People and Bulletin | every 2 min | Shown on the dashboard | 4 Aug |
| Client versions of every validator, and from October every system-chain collator, against the latest polkadot-sdk release | every 6 h | Weekly Tuesday reminder to operators running old versions | 6 Aug (reminder 25 Sep) |
| Validator whitelist on Asset Hub staking: validators not on the list are removed from the validator candidates | every 4 h | Matrix | 22 Jul |
| Bulletin storage used against a 200 GiB threshold | every 15 min | Matrix | 17 Jul |
Two additions came in October:
- the checks now also run over IPv6, so a provider with a broken IPv4 route is not reported as down;
- a check that the number of validators registered on Asset Hub matches the configured validator count.
Funding and payouts
Paseo is funded by a Polkadot treasury bounty held in USDC on Polkadot Asset Hub. Referendum 1919 funded H2 2026 with 210,800 USDC, 46% less than H1. On 29 September the curators closed the H1 bounty (#0), returning its unspent 328 USDC to the treasury, and activated bounty #1.
The Q3 payouts went out on 1 October as 30 child bounties of bounty #1, all in Polkadot Asset Hub block 21,285,403:
- 98,810 USDC paid for Q3, 98.4% of the quarterly budget.
- 111,990 USDC left for Q4 and contingency.
Carried into Q4
- Relay runtime upgrade to v2.5.0-relay, released in Q3 and not yet enacted.