Crossref has turned a future Metadata Plus infrastructure change into a concrete planning date: from 1 February 2027, DOI lookups are due to fall from 150 requests a second per token with no concurrency cap to 64 requests a second with 16 concurrent requests, while search queries are due to fall to 16 requests a second with 8 concurrent requests (Crossref community post). For research offices that ingest Crossref records into local grant, repository or reporting systems, that is not a cosmetic announcement. It defines the throughput budget their existing clients will need to tolerate.
The practical implication is narrow but important. Crossref’s post establishes the new ceilings and the implementation date; it does not establish how any institution will be affected in practice (Crossref community post). Still, the logic is straightforward: any workflow designed around the current 150-per-second allowance or unrestricted parallelism has a testable question before 1 February 2027. Can the same job finish on time under lower per-token rates and explicit concurrency limits? That is an operational check, not a prediction of failure.
Crossref also said it will add monthly Metadata Plus snapshots in Apache Parquet, in beta from February 2027, alongside the existing JSON and XML formats (Crossref community post). That matters because the standing snapshot documentation, last updated on 2020-04-08, still describes a different present-tense world: monthly snapshots served through the REST API’s `/snapshots` route as compressed tar.gz archives in JSON or XML, usually available by the fifth day of the month, covering more than 185 million records, and producing files larger than 200 GB (Metadata Plus snapshots documentation). The safest reading is that Parquet is a planned addition, not a currently documented production default.
The October 2026 funding-data incident makes that distinction more than procedural. Crossref’s status page says that, after a 15 September 2026 change in how funder identifiers were identified, new funder-to-ROR matches were missing from the Metadata Plus REST API pool and from the then-most-recent JSON snapshot on 21 October 2026 (Crossref status). A 24 October update said affected records had been reindexed and that newly updated works in the Metadata Plus pool again included funder-name matches to ROR IDs; the incident was marked resolved on 29 October 2026 at 01:10 UTC, and the dashboard retrieved on 4 November 2026 showed the Plus REST API and JSON Snapshots as operational (Crossref status).
That gap sits awkwardly beside Crossref’s 15 September 2026 blog post, which presented the new matching process as enrichment for new or updated records that included funding information but no funder identifier, while keeping Open Funder Registry identifiers in place and reporting 99% precision in Crossref’s own testing (Crossref blog). The public record therefore supports two propositions at once: Crossref had launched a new enrichment method, and Metadata Plus briefly failed to expose some of those new matches consistently.
For administrators, the useful distinction is between what is announced and what is evidenced. As of 4 November 2026, the record supports a planned 1 February 2027 rate-limit change, a planned February 2027 Parquet beta, and a resolved 21-29 October 2026 Metadata Plus incident (Crossref community post; Crossref status). It does not show measured institutional impact, a finished migration toolkit, or which specific later snapshot fully closed the October data gap. That makes preparation more about verification than assumption: test clients against the published ceilings, treat Parquet as future until production documentation catches up, and recheck October 2026 funder-match-dependent extracts rather than inferring that every downstream copy was complete simply because the incident was closed.
