What were the notable Meta updates?

[Fizz] Do not allow abort reentrancy, [Fizz] Continue reporting aborted task errors after a fatal abort, and [Fizz] model fb bundle's external work scheduling explicitly.

REACT FIZZ TIGHTENS ABORT SEMANTICS, FIXES WORK SCHEDULING VISIBILITY

By RepoJournal · Filed 06:02 UTC on May 31, 2026 · About Meta

React's server-side rendering layer just became harder to misuse, with three commits that eliminate footguns in error handling and work scheduling that could silently swallow failures.

React Fizz is enforcing stricter abort semantics across three coordinated changes landing in the same window. First, abort reentrancy is now explicitly forbidden [1]. Previously you could contrive sequences where onError fired after a render already fataled, creating hard-to-debug timing issues. Now the request state machine rejects any abort attempt on a request that's already aborting, completed, or fataled. Second, when an abort does occur, all pending task errors now report correctly [2]. The old behavior let fatal errors suppress error reporting from tasks inside Suspense boundaries, making it impossible to surface what actually went wrong during an abort sequence. Third, Fizz's external work scheduling is now modeled explicitly in config [3], solving a reasoning problem: you could call scheduleWork with no guarantee it would execute in certain bundle configurations. These changes don't add features. They remove ways the system can lie to you about what happened.

Action items

→ Review your Fizz abort handling and error callbacks for timing-dependent logic facebook/react [plan]
→ If you drive Fizz through external performWork calls, verify your scheduling config after upgrading facebook/react [plan]

References

[1] [Fizz] Do not allow abort reentrancy ↗ facebook/react
[2] [Fizz] Continue reporting aborted task errors after a fatal abort ↗ facebook/react
[3] [Fizz] model fb bundle's external work scheduling explicitly ↗ facebook/react

Quick answers

What shipped in Meta on May 31, 2026?: React's server-side rendering layer just became harder to misuse, with three commits that eliminate footguns in error handling and work scheduling that could silently swallow failures. In total, 3 commits and 3 pull requests landed.
What were the notable Meta updates?: [Fizz] Do not allow abort reentrancy, [Fizz] Continue reporting aborted task errors after a fatal abort, and [Fizz] model fb bundle's external work scheduling explicitly.

REACT FIBER FIXES SINGLETON PROPERTY BLEED, REACT NATIVE STRIPS DEAD CODE

React plugged a critical Fiber bug where singleton cleanup was nuking third-party attributes, while React Native dropped unused timing code from iOS binaries.

javascript 11 shipped 1-min read

Meta 1 day ago

REACT DEVTOOLS BRIDGE HARDENS AGAINST RECONNECTS AND CORRUPTION

React's DevTools layer just got bulletproof: hoxyq landed four interconnected patches that eliminate silent state corruption, validate invariants before mutation, and handle extension reconnects with message buffering and handshakes.

javascript 22 shipped 2-min read

Meta 2 days ago

REACT SERVER DOM TURBOPACK GETS DYNAMIC CHUNK MERGING

React's Turbopack integration now supports an experimental array format that lets you dynamically merge or unmerge chunks based on what's already loaded in the browser.

javascript 16 shipped 1-min read

Meta 3 days ago

React 19 Cuts Server Component Decoding Time, React Native Fixes requestIdleCallback Regression

React shipped performance wins across 19.0.8 and 19.1.9 while React Native patched a critical timeout bug that broke requestIdleCallback tasks.

javascript 22 shipped 2-min read

Elsewhere on the wire

AI Agents about 10 hours ago

CLAUDE OPUS 5 LANDS ACROSS THE STACK

The newest Anthropic model is now live in langchain, Cline, and llama-index, with native support for extended reasoning and 1M context windows.

ai-agents 28 shipped 1-min read

Local LLMs about 10 hours ago

OLLAMA LANDS LAGUNA SUPPORT AND CRUSHES MEMORY LEAKS WHILE SGLANG HITS V0.5.16 WITH CONFIDENCE-DRIVEN SPECULATIVE DECODING

Ollama shipped three critical performance and reliability fixes for Metal residency and concurrent access patterns, while SGL-Lang released 0.5.16 with a new speculative algorithm hitting 383.7 tok/s on DeepSeek-V4.