Switchboard’s September 25 support deadline has turned a six-day migration warning into a test of Solana’s price feeds. Current public documentation shows where its data remains part of an application’s design, but those pages cannot prove that a live market is still using the feed. Jito and marginfi provide two sharply different views of the exposure.

Switchboard has reached its stated September 25 end of technical support, leaving Solana applications to verify the price sources configured in their live programs.

The oracle project’s September 19 statement, as reproduced in coverage of the announcement, said its core development contributor Switchboard Technology Labs would wind down and all implementations were deprecated immediately. The team urged integrators to migrate to other providers, naming Pyth and RedStone. September 25 was described as the last day for existing support. A company ending support is a real operational milestone. It does not, by itself, prove that every onchain feed stopped updating at midnight or that every application once associated with Switchboard remained dependent on it.

Switchboard’s own documentation has named Kamino, Jito, marginfi and Drift as users. Those are historical integration claims from a provider that was selling an oracle service, not a real-time inventory of active feeds on September 25. Checking each project’s current documentation reveals a more complicated picture. Jito’s Tip Router pages still describe Switchboard in their pricing flow; marginfi’s September technical upgrade adds paths designed to avoid that dependency. One document can be stale while another anticipates a migration. Neither substitutes for an inspection of live account configuration.

The earlier Switchboard funding round was $7.5 million in May 2024. The amount is useful background on the venture’s history, but it gives no measure of today’s protocol exposure. The relevant count is the number and value of live markets whose risk calculations still take data from a feed that cannot be reliably updated, and that count cannot be inferred from a customer logo.

A listed integration is not an active feed

Switchboard’s public introduction describes on-demand feeds: applications create or call the data they need, and a price is made available through Solana accounts. Documentation can identify where a protocol knows how to read a Switchboard feed. It may not identify which option a particular market currently selects. A software development kit may support an oracle type long after the last bank switches away from it. Conversely, a website may change while a live reserve retains its older oracle account.

Three levels of evidence need to be kept apart. First is a marketing or integration page, which shows a relationship existed. Second is a program’s supported configuration, visible in technical documentation or code. Third is the live configuration and recent update history of the actual market. Only the third can support a claim that a named market still relied on Switchboard at a given time. Even then, a backup source may be configured, so the impact of a stopped primary feed must be checked against the relevant fallback and freshness rule.

Consider marginfi’s protocol documentation. Its oracle table retains SwitchboardPull and venue variants among available setups. It says a caller must crank a Switchboard pull feed just before use. The same table lists Pyth push feeds and Scope accounts as other setups. A reader could mistake the continued Switchboard row for proof that every marginfi bank still uses it. The table describes supported types, not a complete list of which bank uses which feed today.

BREAKING: Pyth price feeds now live in Etherfi’s Earn product on Optimism, powering the largest crypto neobank with over $5.7B TVL pic.twitter.com/Yvtoc97hvM

— crypto.news (@cryptodotnews) April 17, 2026

Marginfi’s separate Program 0.1.11 note is more current and more specific. It instructed developers to upgrade the SDK to at least version 2.8.0 before September 4, saying banks would begin moving to new oracle setups from that date. The release added nine variants that do not depend on Switchboard, including Kamino Scope feeds and exchange-rate based pricing for certain liquid staking and principal tokens. The note does not say every bank had migrated by September 25. It does show that a project publicly documented a route away from the threatened dependency before the shutdown announcement.

The migration carries a surprising second failure mode. Marginfi says older SDKs cannot decode a bank configured with one of the new oracle enum values. A single bank with an unsupported value can prevent Project0Client.initialize and bank reads, not merely an action involving that bank. In other words, changing an oracle can fix one infrastructure dependency while breaking an integrator that has not updated its software. Marginfi’s document tells integrators how to avoid the SDK problem; it is not evidence that any particular user suffered it.

Project 0 has described unified margin across Solana venues, including Kamino and Drift. Cross-protocol interfaces create another layer at which an oracle migration must be read correctly. The note about older SDK versions is concrete evidence of an integration hazard, without proving a failure in Project 0 or any other named app. A responsible audit would check software versions and live lending-bank configurations before claiming an outage.

Jito’s Tip Router still documents Switchboard

Jito Foundation’s Tip Router overview says Switchboard determines the relative weight of assets such as JitoSOL and JTO held in vaults linked to the Tip Router. The overview identifies an onchain Tip Router program, a node-operator client and a permissionless cranker. Its pricing documentation names Switchboard as the current oracle feed and describes backup weights when feeds are unavailable.

The documents place Switchboard in a specific job: pricing vault assets for weight calculations in a tip distribution and restaking system. They do not say that an unavailable Switchboard feed would automatically liquidate a Solana lending position. Jito’s pricing page describes a fallback mechanism, which weakens the simplistic claim that a support sunset necessarily makes all Tip Router operations stop. The exact fallback values, activation conditions and current live oracle accounts still need a current program-state check.

The Tip Router overview showed a last-updated marker of nine months ago when checked on September 25. That age changes how it can be used. It establishes a documented design and identifies where to ask a technical question. It cannot establish that the present program has the same feed configuration. Jito may have updated onchain accounts without revising the page, or it may still use Switchboard with a fallback. Without a recent transaction inspection or a current statement from Jito, a named live dependency remains unverified.

Jito’s public GitHub release notes for Tip Router refer to retrying Switchboard oracle gateways in keeper operations. A codebase containing such logic likewise demonstrates technical integration, not necessarily a dependency of every vault at the time of publication. Code can preserve a compatibility path for months. The live question is whether recent price update transactions target a Switchboard account used by a vault still carrying value, and whether that account advances after the support deadline.

The distinction is often lost when all oracle users are placed in a single list. Jito’s described calculation affects relative asset weights in a distribution system. A lending market’s described calculation determines collateral value and borrower health. Both consume price data, but their failure paths differ. An audit that counts logos would assign the same severity to fundamentally different uses.

Kamino’s Scope is an aggregator, not a provider label

Kamino Finance’s public Scope repository describes an onchain aggregator that copies values from multiple oracle accounts into one price feed and validates updates under preset rules. Its README says a feed supports up to 512 prices and that the association between an index and a token pair is not wholly stored onchain. A downstream program may point to Scope while Scope itself relies on other feeds for the selected asset. Seeing Scope in a bank configuration is thus a starting point for tracing the actual data source, not the end.

The September marginfi note lists Scope as an option that does not depend on Switchboard for the new setup it describes. That does not imply every deployment of Scope on every date excludes every Switchboard source. An aggregator can change its underlying inputs. A complete dependency check needs both the consumer’s selected Scope account and the source mapping used to populate its entry. Kamino’s repository supplies the architecture, not a timestamped inventory of current mainnet sources for every application.

Kamino has continued bringing institutions into its lending ecosystem. Galaxy opened two stablecoin vaults on the platform in September. The existence of new vaults shows why naming a whole protocol as exposed without checking its individual assets would be unsound. A USDC vault, a liquid staking token reserve and a tokenized equity market can use different oracle paths. We have not verified that Galaxy’s vaults use Switchboard, so they are not included in a count of affected positions.

Similarly, the older list of Kamino, Jito, marginfi and Drift in Switchboard’s introductory material does not tell us the distribution of exposure among them. A project may use an oracle only for one market, use it as a fallback, or retain code after switching live feeds. The only defensible unit of analysis is a specific market or vault and its configured feed at a specified time. Without that unit, claims about funds at risk are marketing arithmetic run backward.

A stale feed has more than one possible effect

The technical consequence of a feed falling behind depends on the consuming protocol. A lending program generally needs a price to determine collateral value and borrowing capacity. If it rejects an old value, an action may fail or a market may pause under its rules. If it accepts stale data, a borrower might transact against a price that no longer matches the market. A fallback source may keep the market operating but introduce a new update rhythm or confidence rule. The protocol’s documentation and onchain configuration decide which path applies.

Marginfi explicitly says Switchboard pull feeds need to be cranked before use. An integrator must therefore supply a fresh update as part of its transaction path. Pyth push feeds, by contrast, are described as being kept fresh through Pyth’s infrastructure. Scope uses an aggregated account value selected by a configured entry index. Moving between these types changes the accounts a transaction needs and the code that checks them. The September SDK warning is one visible example of those changes reaching application software.

For Jito Tip Router, the public docs describe backup weights for unavailable feeds. Whether those backups preserve accurate reward allocation through a sustained outage is a question for live configuration and Jito’s operators, not something a documentation sentence resolves. If a feed keeps updating through independent node operators after the company stops support, no fallback may be triggered immediately. If updates cease but the backup is active, operations may continue with a different pricing method. These are conditional paths, not a prediction of the system’s present state.

An unrelated oracle incident led to liquidations on Vesu earlier in September. It illustrates that incorrect pricing can have economic effects, but it is not evidence of an incident at Switchboard, Jito or marginfi. A shutdown notice should not be turned into a liquidation claim by analogy. The sign of an actual event would be stale account timestamps, failed transactions, a protocol pause or identified losses, none of which has been shown here for the September 25 deadline.

Solana’s move to 250 millisecond slots changed the pace at which blocks are produced, but it did not guarantee that an external price source updates. Faster slots can carry a new price sooner when one exists. They cannot manufacture a price when the node supplying it stops. A protocol’s freshness test may be measured by slot, time or another rule, so a change in the network clock may alter how developers interpret old feed configurations.

Who bears the migration work?

The oracle operator publishes or coordinates data, but the consuming protocol chooses the account its program reads and the limits it places on that price. A lending protocol can require governance or an administrator to change oracle addresses for its markets. Its front end and third-party integrators then have to construct transactions with the right additional accounts. Users may only notice a rejected borrow or a paused market, long after the operator and protocol have made their technical decisions.

An operator ending support does not necessarily have the power to rewrite a customer’s program configuration. The Switchboard notice urged users to migrate because integration owners must act. Projects should be assessed by the addresses and account updates they control. If an application already moved to Pyth before September 19, the later support deadline has no direct effect on that market. If it still selects a Switchboard feed and has no working backup, the feed’s behavior after September 25 is the concrete issue.

BREAKING: Pyth becomes an external distributor of Nasdaq Basic

The partnership brings Nasdaq’s real-time U.S. equity quotes and trade data to the Pyth Data Marketplace, expanding access to institutional-grade market information across blockchain applications. pic.twitter.com/7sIrpWL1pI

— crypto.news (@cryptodotnews) September 22, 2026

The strongest opposing reading of the shutdown alarm follows from marginfi’s own September note and Jito’s documented backup. Applications can design redundancy or move ahead of a vendor exit; the code and documents show mechanisms for doing so. Switchboard’s on-demand model can leave some feed infrastructure running independently even if the core contributor has stopped support. The notice did not publish a verified schedule at which every account would halt, and we found no primary evidence establishing such a universal cutoff.

There is a different kind of continuity question for a protocol that made its own fallback. A backup price can prevent a total stop while pricing an asset less frequently or with a different source set. For a reward distribution process, a temporary backup weight may keep epoch accounting moving, although the allocation may then rely on the backup’s assumptions. For a lending market, the fallback could change the price used in a health check. These are not claims about current Jito or marginfi settings. They show what a maintainer must disclose before users can judge whether a migration is complete in operational terms, not merely whether transactions still execute.

A provider wind-down can have delayed effects as well. Code written to request on-demand prices may succeed while an independent gateway answers, then fail when that gateway is retired or its operators stop updating a specific asset. An observer needs several post-deadline timestamps, not a single successful transaction, to infer continued service. The same discipline applies to a failed transaction: one user’s error may arise from a stale SDK or insufficient account input instead of an unavailable oracle. Marginfi’s migration document supplies an explicit example of a software decoding failure that could otherwise be mislabeled as an oracle outage.

JUST IN: Hyperliquid SK Hynix perpetual drops 17.9% on oracle anomaly

Forced liquidations followed an erroneous external market print relayed by the pricing oracle pic.twitter.com/X2YYeLRFiJ

— crypto.news (@cryptodotnews) July 29, 2026

There is a limit to that reassurance. A fallback described nine months earlier needs validation against current state, and a migration option described in September is not proof every bank took it. The two documents supply credible reasons not to assume catastrophe, while leaving a measurable gap. The fair conclusion is narrower than both the promotional and alarmist versions: public documents identify candidate dependencies and escape routes; a current market-by-market configuration audit is needed to establish any remaining exposure.

The live inventory is still the missing document

The original reporting here compares Switchboard’s list of four prominent integrators with current primary documents from Jito, marginfi and Kamino. It yields two verified documentary findings. Jito’s older Tip Router documentation names Switchboard for vault pricing and a fallback for unavailable feeds. Marginfi’s September 0.1.11 note describes nine new setups independent of Switchboard and warns of a separate SDK break if integrators do not upgrade. Kamino’s Scope repository explains why an aggregator label alone cannot identify every upstream data source.

The work does not produce a count of live unmigrated feeds, user funds exposed or an outage at any named protocol. The available public pages do not contain a synchronized September 25 snapshot of all oracle accounts, latest successful updates, fallback settings and amounts supported by each market. Claiming a specific dollar total from protocol TVL would be indefensible, because the whole protocol’s assets do not necessarily share the same oracle. The precise headline question remains open at the live-account level.

A proper count would use the market as the row, not the protocol. For each active lending bank, derivative market or reward vault, the auditor would record its program address, selected oracle type, oracle account, backup source if any, latest successful price update, maximum permitted age and the value of positions actually dependent on that particular price. Duplicate markets that share one oracle account should not be counted as distinct feeds; one market using two independent oracles should not be counted as wholly dependent on either without reading its fallback logic. The timestamp of the market configuration matters because an administrator could change a feed after the observation.

This method explains why even a true statement such as a protocol supported 550 feeds in the past is insufficient for the present question. A feed can exist without an active borrower, can have a price update without a consuming market, or can be referenced only in dormant code. A count of feed accounts measures infrastructure. A count of configured markets measures dependency. A count of positions and collateral actually touching those markets measures economic exposure. None is interchangeable with total assets deposited in all products run by a project.

There is a further verification step when a source is an aggregator. The consumer may identify a Scope account and entry index, while the Scope mapping points onward to one or more providers. An update in the Scope account after September 25 proves an aggregator produced a value, but it does not by itself prove Switchboard continued to supply the underlying price. The investigator needs the selected entry and source configuration for that update. Kamino’s repository notes that token-pair labels are not entirely stored onchain, so external configuration or maintainer documentation may be needed to map an index to its asset. Where that mapping is unavailable, the result should be recorded as unknown, not silently attributed to Pyth or Switchboard.

What to watch

  • Market oracle addresses: Compare each active bank or vault’s configured feed with the documented Switchboard accounts.
  • Price update timestamps: Check whether an identified feed continues publishing fresh values after September 25.
  • Fallback configuration: Look for the source and freshness limit used if a primary feed falls behind.
  • Recent program transactions: Check whether borrowing, settlement or tip distribution still completes for the affected market.
  • Dated maintainer updates: Look for a named migration, market pause or remaining dependency, supported by an account or program address.

Record the observation time for each check; a screenshot without a block or timestamp can quickly become stale.

Marginfi’s upgrade note states that a bank using a new oracle enum value can make an older SDK fail to initialize its client, even if a user does not interact with that particular bank. The instruction to use SDK version 2.8.0 or later was published ahead of September 4’s migration start, three weeks before Switchboard’s support deadline.

FAQ

When did Switchboard say support would end?

The shutdown announcement was made on September 19, 2026, and identified September 25 as the end of existing technical support. The notice deprecated implementations immediately.

Did all Switchboard oracle feeds stop on September 25?

The support deadline alone does not establish that every onchain account stopped updating. Current transaction and feed timestamps are needed to make that claim.

Does Jito still use Switchboard?

Jito’s Tip Router documentation still names Switchboard in vault pricing, but its overview is marked as last updated nine months earlier. The pages do not prove the live September 25 configuration.

Did marginfi migrate off Switchboard?

Marginfi’s September upgrade documents nine new oracle setups that do not depend on Switchboard and says banks began moving from September 4. It does not state that every bank completed a migration.

Why can an oracle migration break an SDK?

Marginfi says older SDKs do not recognize the enum values used by its nine new setups. A bank configured with one can make an old client’s initialization fail; version 2.8.0 or later supports the variants.

Is Kamino Scope independent of every external oracle?

Scope aggregates values from other oracle accounts. Its presence in a consumer’s configuration does not identify every upstream source without examining the specific entry mapping.

How can users check whether a market is affected?

The market’s configured oracle account, latest update and fallback settings provide a stronger answer than a historical provider list. Protocol announcements can confirm whether a specific market has migrated.

Have losses been verified from this shutdown?

No losses at a named protocol were verified for this feature. An earlier incident at another protocol cannot prove one occurred here. This is educational analysis, not investment advice.

Disclaimer: This article is for information and educational purposes only and does not constitute financial or investment advice. Figures reflect regulatory filings and reporting available at the time of writing and change with each disclosure. Nothing here is a recommendation to buy, sell, or hold any security or asset. Always do your own research. Information is accurate as of September 25, 2026.