| Report Date | Date of submission (2026/06/04) |
| Submitted by | Andrei Sandu |
This period continued my multi-quarter effort to improve parachain block production reliability. My central focus was block confidence work — diagnosing the root causes of low block confidence and mitigating them by enabling longer AURA slots and faster block propagation. I also supported the Bulletin Chain launch with integration and stress testing of 6s, 2s and 1s block configurations and authored a high-level proposal for Polkadot SDK release QA automation. I started the cleanup of legacy collation/validation protocol code, and acted as a reviewer and design consultant for the low-latency parachain work. Beyond that, I delivered a small protocol improvement raising the code-size limits required to support larger parachain code upgrades.
I drove the diagnosis and resolution of the issues affecting block confidence in production chains AH and PC.
Investigation and root-cause analysis. I drafted the plan to deploy the YAP canary parachain on Polkadot (#10933, runtime in #10165) running 12 cores and 500ms blocks. From this I identified and documented the dominant causes (#10860), the gap between Polkadot Asset Hub and YAP/Kusama confidence (#11827), and slow availability where the block producer does not see 2/3 of bitfields in time (#11582).
Longer AURA slots. I migrated parachains to 24s AURA slots — across all testnet parachains (#11778), the Bulletin Chain (runtimes#1149), and Kusama/Polkadot Asset Hub and People Chain (runtimes#1174). With a longer slot a single collator authors several consecutive blocks, which reduces slot-boundary fork contention between collators (improving confidence) and improving Elastic Scaling throughput.
Block propagation I investigated AH and concluded that freshly built block traverses multiple network hops to reach the next author, and each hop costs roughly 3x the inter-peer latency (announce, request, response). Late blocks force authors to build on stale parents resulting in forks.
I took the following actions to tackle this issue:
I raised the RC block-size and code-size limits required to support larger parachain code upgrades, enabling parachain code size up to 5MiB (#11894, #11893), with the corresponding production runtime change in runtimes#1157.
I fixed Fellowship zombienet test flakiness, extended the automated test coverage for Elastic Scaling, and supported the Bulletin Chain launch (stress testing, stress test tool improvements and investigating node issues).
I authored a high-level design for automating Polkadot SDK release QA
(#12054). I identified the gaps in our
current release process. Release candidates are validated manually with no real
application load, no long-duration runs, no mixed-version (matrix) testing, no negative scenarios,
and no verification against production runtimes. I proposed a dedicated, scalable test environment
created by forking test/prod networks with zombie-bite, triggered by SDK release and Fellowship
runtime releases. The goal is to mimic real-world conditions (multiple node versions, simulated
latencies and faults, malicious actors, simulated user load) and runs continuously for days at a
time, with results and individual failures aggregated and reported automatically. The proposal
defines an MVP focused on block production, block confidence, and finality for SDK releases.
I scoped and drafted the retirement of legacy code paths to reduce maintenance burden and attack
surface: removing v1 CandidateDescriptor support
(#11373)
and simplifying the upgrade path of the validation and collation protocols
(#11412). The implementation work —
disabling v1 candidate receipts (#11397) and
removing v1 support from the collator protocol (#11414)
— is drafted but currently parked, since I must sync with parachain teams that might still use old
versions.
I acted as a consultant and reviewer for the low-latency parachain design and its implementation,
contributing to the design discussions and reviewing the PRs that build it — notably the support
for older relay parents, relay-parent advertisement in v3, v3-descriptor scheduling and v3 Cumulus
support.
I also reviewed the ongoing collator-protocol revamp
(#8541,
#11046,
#10917 and related PRs). In total I reviewed
70 merged PRs in polkadot-sdk during the reporting period.
A full list of merged contributions in polkadot-sdk for the period is available
here.
| Ranks | Activity thresholds | Agreement thresholds | Member's voting activities | Comments |
|---|---|---|---|---|
| III | 70% | 100% | 100% | 47 out of the 47 |
Question(s):
Concern(s):
Comment(s):