[SandroMaglione]

Summary

• A local MachineRef.send to an idle machine now processes the event in the sender's turn. Work that completes without suspending, including raised events and the committed state, is reflected by snapshot when send returns. Suspending work, such as an invocation, continues in the background. Events sent while the machine is busy are queued, as before. MachineTarget.send, and so any future Cluster delivery, keeps promising only delivery.
• Controlled inputs bound to machine state through AtomMachine now keep their value. Before, React restored the previous value because the state arrived a turn later, which also left stale form-validation messages after autofill. Local references expose an internal synchronous snapshot observer that the atom bridge uses; references without one fall back to changes.
• Performance:
  • Machines without invocations drain synchronously, without a worker fiber, and keep their compact execution state while idle.
  • changes publication is coalesced through the sender fiber's scheduler dispatcher, as Queue releases its takers, so a burst still reaches subscribers in one publication.
• The generic reference runtime now uses a mailbox that wakes a waiting receiver inline, so both strategies share delivery timing. It also publishes EventSent before the receiver processes the event.
• Fix: stopping an idle machine now completes an observed emissions stream; it previously never ended.

Tests:

• Public: test/machine/SynchronousDelivery.test.ts, covering machines with and without invocations, raised events, emissions, observer re-entrancy, and suspended invocations.
• Generic-versus-compiled differential coverage, including targetless, reentering, terminal, raised-event, active-invocation and retained-snapshot cases.
• AtomMachine tests for root and invoked-child projections.
• A jsdom controlled-input test in effect-machine-react.

Changeset

• Added or updated for a library or package-metadata change (minor: the timing of send is observable)
• Not required because this PR does not change a publishable package

Validation

• pnpm check
• Automated type-performance measurement passed or was not required (pnpm perf:types passes the budgets locally; only internal types and documentation changed)
• Automated runtime- and memory-performance measurement passed or was not required (pending the workflow; the change touches the send hot path)

Also checked against the downstream app that reported the bug, with Playwright:

• Typing in the middle of a field and autofill after typing now pass on every run.
• The "autofill on load" scenario still fails when hydration is slow. React resets the input during hydration before any handler runs, and the original 0.44.0 build shows the same result, so this PR does not address it.

🤖 Generated with Claude Code

[github-actions]

Type performance

Measured with TypeScript 6.0.3 and skipLibCheck=true.

Scenario	Base	PR	Difference
Effect only	55	55	0 (0.0%)
Import effect-machine	55	55	0 (0.0%)
Machine.state (3 children)	2,664	2,664	0 (0.0%)
Machine.make (3 states, 2 events)	8,932	8,932	0 (0.0%)
machine.handle (3 states, 2 transitions)	39,449	39,449	0 (0.0%)
declared transition (10 named branches)	60,402	60,402	0 (0.0%)
atomic construction, guards, and reentry	85,710	85,710	0 (0.0%)
registered invocation (input-taking Effect)	39,819	39,819	0 (0.0%)
machine.handle (depth 24)	194,531	194,531	0 (0.0%)
machine.handle (wide depth 16)	218,238	218,238	0 (0.0%)
machine.handle (parallel/history/choice)	91,642	91,642	0 (0.0%)
machine definition (3 independent implementations)	99,173	99,173	0 (0.0%)
machine exact input/output/error/services	82,941	82,941	0 (0.0%)
execution adapter readiness	71,428	71,428	0 (0.0%)

Marginal instantiations are measured against the matching setup without that API call:

Scenario	Base	PR	Difference
Import effect-machine	0	0	0
Machine.state (3 children)	2,609	2,609	0 (0.0%)
Machine.make (3 states, 2 events)	6,260	6,260	0 (0.0%)
machine.handle (3 states, 2 transitions)	30,517	30,517	0 (0.0%)
declared transition (10 named branches)	51,849	51,849	0 (0.0%)
atomic construction, guards, and reentry	70,370	70,370	0 (0.0%)
registered invocation (input-taking Effect)	32,682	32,682	0 (0.0%)
machine.handle (depth 24)	177,222	177,222	0 (0.0%)
machine.handle (wide depth 16)	201,707	201,707	0 (0.0%)
machine.handle (parallel/history/choice)	79,144	79,144	0 (0.0%)
machine definition (3 independent implementations)	88,492	88,492	0 (0.0%)
machine exact input/output/error/services	70,732	70,732	0 (0.0%)
execution adapter readiness	31,922	31,922	0 (0.0%)

Check times (informational)

Scenario	Base	PR
Effect only	0.03s	0.03s
Import effect-machine	0.03s	0.03s
Machine.state (3 children)	0.09s	0.09s
Machine.make (3 states, 2 events)	0.14s	0.14s
machine.handle (3 states, 2 transitions)	0.30s	0.30s
declared transition (10 named branches)	0.38s	0.38s
atomic construction, guards, and reentry	0.51s	0.50s
registered invocation (input-taking Effect)	0.31s	0.31s
machine.handle (depth 24)	0.63s	0.64s
machine.handle (wide depth 16)	0.71s	0.70s
machine.handle (parallel/history/choice)	0.51s	0.52s
machine definition (3 independent implementations)	0.53s	0.54s
machine exact input/output/error/services	0.51s	0.49s
execution adapter readiness	0.45s	0.43s

Type instantiations are the comparison metric. Check time varies with runner load and is informational only.

[github-actions]

Runtime performance

Median of 5 independent benchmark processes on AMD EPYC 7763 64-Core Processor with Node v24.21.0.

Pull request baseline

Scenario	Effect Machine
Plan counter transitions	112,739 transitions/s
Drain burst with terminal fence	400,825 increments/s
Drain burst with a change observer	374,012 increments/s
Lookup and send to one child	310,125 increments/s
Start and stop a machine	115,929 machines/s
Start and stop a parent with one child	23,234 families/s
Complete an Effect invocation with tracing enabled	7,260 invocations/s
Complete an Effect invocation with tracing disabled	8,745 invocations/s
Plan transitions through a compound state	101,287 transitions/s
Plan transitions through parallel regions	80,319 transitions/s
Drain burst through a compound state	343,261 events/s
Drain burst through two parallel regions	318,828 events/s
Drain a compound-state burst with a change observer	319,948 events/s

Process runtime reference points

Scenario	Effect Machine
Start and stop a raw generic process	20,842 processes/s
Start and stop a raw compiled process	68,985 processes/s

Memory profile	Effect Machine
Idle machine	1.9 KiB
Raw generic managed process	11.1 KiB
Raw compiled process	3.1 KiB
Two independent idle machines	3.5 KiB
Idle parent with one child	6.0 KiB
Parent with observed child registry	9.5 KiB
Parent with observed invoked child snapshots	6.9 KiB

Effect Machine change from base

Metric	Base	Base variability	PR	PR variability	Difference
Plan counter transitions	113,455 transitions/s	0.4% MAD	112,739 transitions/s	1.1% MAD	-0.6%
Drain burst with terminal fence	396,551 increments/s	1.1% MAD	400,825 increments/s	0.8% MAD	+1.1%
Drain burst with a change observer	368,462 increments/s	1.4% MAD	374,012 increments/s	0.4% MAD	+1.5%
Lookup and send to one child	323,565 increments/s	0.4% MAD	310,125 increments/s	0.4% MAD	-4.2%
Start and stop a machine	114,077 machines/s	2.7% MAD	115,929 machines/s	2.0% MAD	+1.6%
Start and stop a parent with one child	23,474 families/s	2.5% MAD	23,234 families/s	1.2% MAD	-1.0%
Complete an Effect invocation with tracing enabled	7,000 invocations/s	0.5% MAD	7,260 invocations/s	1.0% MAD	+3.7%
Complete an Effect invocation with tracing disabled	8,490 invocations/s	1.3% MAD	8,745 invocations/s	1.6% MAD	+3.0%
Plan transitions through a compound state	99,721 transitions/s	0.4% MAD	101,287 transitions/s	1.0% MAD	+1.6%
Plan transitions through parallel regions	79,416 transitions/s	0.7% MAD	80,319 transitions/s	2.2% MAD	+1.1%
Drain burst through a compound state	337,390 events/s	1.3% MAD	343,261 events/s	1.4% MAD	+1.7%
Drain burst through two parallel regions	312,225 events/s	0.3% MAD	318,828 events/s	1.8% MAD	+2.1%
Drain a compound-state burst with a change observer	316,426 events/s	1.3% MAD	319,948 events/s	1.2% MAD	+1.1%
Idle machine heap per unit	1.9 KiB	0.5% MAD	1.9 KiB	1.3% MAD	-0.4%
Raw generic managed process heap per unit	11.7 KiB	0.0% MAD	11.1 KiB	0.1% MAD	-5.3%
Raw compiled process heap per unit	3.1 KiB	0.0% MAD	3.1 KiB	0.0% MAD	+0.6%
Two independent idle machines heap per unit	3.5 KiB	0.0% MAD	3.5 KiB	0.0% MAD	+0.9%
Idle parent with one child heap per unit	6.0 KiB	0.0% MAD	6.0 KiB	0.0% MAD	+0.5%
Parent with observed child registry heap per unit	9.5 KiB	0.0% MAD	9.5 KiB	0.0% MAD	+0.3%
Parent with observed invoked child snapshots heap per unit	6.9 KiB	0.0% MAD	6.9 KiB	0.1% MAD	+0.4%

Process runtime reference change from base

Metric	Base	Base variability	PR	PR variability	Difference
Start and stop a raw generic process	19,617 processes/s	2.0% MAD	20,842 processes/s	0.9% MAD	+6.2%
Start and stop a raw compiled process	67,123 processes/s	0.8% MAD	68,985 processes/s	0.4% MAD	+2.8%

Regression guard

No large, noise-adjusted throughput or heap regressions detected.

Versions and interpretation

• Effect Machine: 0.44.0

Higher throughput is better; lower heap is better. Variability is the median absolute deviation across independent processes, relative to their median. Small differences on shared GitHub-hosted hardware remain informational; the required guard rejects only large changes beyond the measured noise allowance.