Cardano released node software version 11.2.0 on October 28, 2026, advancing preparations for its Dijkstra upgrade while keeping the build outside the mainnet recommendation. The release gives infrastructure teams software to test, but the project's own test results show why availability is only one step toward deployment.
Intersect, the organization coordinating Cardano development, said on October 30 that the version would move DijkstraNet, a public testing network, into the new era after its stake pool operators upgraded. It described the package as carrying Dijkstra's features apart from Leios, the separate scaling component planned for node 11.3.
For Cardano and ADA, the immediate development concerns network readiness. It does not establish that Dijkstra has activated on the production blockchain.
Configuration changes reach beyond the node itself
The new release changes how a node reads its operating settings. Its published network configurations use a structured envelope, a format older node versions cannot read. Version 11.2.0 still accepts the previous flat format, so compatibility works in one direction rather than both.
The underlying cardano-config library documents separate fields for the format version, the expected minimum node version and the configuration itself. It also supplies migration and checking tools. Migration reshapes an older file while preserving its values; a subsequent check establishes whether the resulting file can be parsed.
That makes configuration review a practical part of testing. Copying a newly published settings file into an older installation is not equivalent to upgrading the executable. Conversely, successfully converting a file does not establish that every connected application is ready.
Wallets and other services have their own compatibility boundary. Version 11.2.0 removes node-to-client protocol versions older than NodeToClientV_23. Software that talks directly to the node therefore needs compatible network libraries. This affects infrastructure integration, distinct from an individual user's wallet custody and recovery arrangements.
Successful Conway tests do not clear Dijkstra
The published release-testing report separates checks under Conway, Cardano's existing era, from work on Dijkstra. Its two Conway regression configurations each recorded 1,908 passing tests, nine known broken tests and 359 skipped tests, with no failures.
Those figures are not a clean bill of health for the next era. Dijkstra testing remained unsuccessful because of two issues: missing local submission support for deliberately invalid second-stage transactions, and a connection closing when a guard signature was rejected. The report points to a likely fix for the latter, without presenting a successful Dijkstra rerun.
The upgrade test also stayed at protocol version 11 when moving from node 11.1.3 to 11.2.0. It tested a software replacement and a period of mixed node versions, not the Dijkstra hard fork. Keeping those test categories separate prevents a passed installation check from being mistaken for completed protocol validation.
A slower synchronization result has limits
A separate mainnet synchronization report, covering an 11.2.0 run on October 28-30, found higher memory consumption and longer synchronization than the selected 11.1.1 comparison run. Peak process memory rose from 21.16 to 22.08 GiB, while total synchronization time increased by 19.4%.
The same report cautions against assigning those differences entirely to the new code. The runs used different consensus modes and snapshot-delay settings, and the newer run synchronized a longer chain. These are measurements from particular test environments, not a prediction of every operator's experience.
There was also a narrower positive finding: the severe repeating memory spikes seen in an earlier version did not return. Memory nevertheless stayed elevated. Reporting both observations matters because eliminating one failure pattern does not establish that the remaining resource issue has been solved.
Leios development continues on a separate track
The consensus team's November 3 update described further Leios implementation and prototype work, including checks on votes and oversized data. Its latest prototype added adaptive endorsements, reducing the size of proposed transaction batches when they fail to attract enough votes. These are development milestones, not evidence that those capabilities are already active on mainnet.
Cardano's official upgrade documentation distinguishes installing node software from activating protocol rules. Operators install compatible software ahead of the switch; a hard fork governance action authorizes the transition on chain. This is part of the coordination required by a proof-of-stake network, rather than an automatic consequence of publishing a release.
The next material evidence will be resolved Dijkstra test issues, a mainnet-ready recommendation and the relevant activation decisions. The October 28 release moves integration work forward. Its pre-release designation and the published test limitations define how far that progress currently goes.