smite-ir: generate valid channel parameters in open_channel flows

[NishantBansal2003]

Closes: #224

[erickcestari]

Should we be very strict and only create valid funding/open_channel flows? Or still allow generators to create invalid cases?

cc @morehouse

[morehouse]

This is a big improvement. I did an experiment with 2k seeds generated from the OpenChannelGenerator. Here's the per-impl accept rates:

Revision	LDK	LND	CLN	Eclair
Before this PR (de64777)	0.00%	0.00%	0.00%	0.00%
Current PR (eb75402)	0.10%	0.00%	6.40%	0.25%

I expect this alone will have a measurable improvement in AUC on a 24h+ experiment, and we should do that experiment to make sure.

We may want to make some further improvements first though. Ignoring channel type and shutdown script (which still cause most rejections even with our explicit enumeration of variants), the remaining top rejection reasons for the channel parameters on the 2k experiment were:

97.98%  LDK dust_limit not in [354, 546]
95.12%  Eclair reserve > 5% of funding
92.62%  LND dust_limit not in [354, 1062]
80.27%  LND channel_reserve > funding/5
80.12%  LND max_in_flight < 5*htlc_minimum
51.80%  Eclair dust_limit not in [354, 5000]
26.29%  CLN htlc_minimum > capacity
25.03%  LDK htlc_minimum >= (funding - reserve) * 1000
24.94%  Eclair to_local and to_remote < reserve
24.94%  LDK push_msat > (funding - reserve) * 1000
23.70%  LDK SCID alias on announced channel
 4.40%  LND max_accepted_htlcs not in [5, 483]

[morehouse]

Should we be very strict and only create valid funding/open_channel flows? Or still allow generators to create invalid cases?

I don't think we can get valid flows 100% of the time without overcomplicating the generator. But it would be good to get single-digit percentages of flows validating successfully, which I think should be possible by addressing some of the remaining blockers flagged above.

[NishantBansal2003]

I expect this alone will have a measurable improvement in AUC on a 24h+ experiment, and we should do that experiment to make sure.

Updated the bounds. I'm currently running the experiment and will share the results shortly

[morehouse]

Code LGTM.

Updated 2k experiment results:

Revision	LDK	LND	CLN	Eclair
Before this PR (de64777)	0.00%	0.00%	0.00%	0.00%
Previous PR (eb75402)	0.10%	0.00%	6.40%	0.25%
Current PR (199dd8d)	8.10%	4.40%	9.95%	8.30%

[NishantBansal2003]

Fuzzing Evaluation Report

Configuration A (Baseline): baseline (Commit: 8474544 (2026-09-04))
Configuration B (Experimental): experimental (Commit: 199dd8d (2026-09-05))

1. Summary Statistics

Target	Duration (h)	n (Baseline)	n (Exp.)	Median Cov. (Baseline)	Median Cov. (Exp.)	Adj. p-value (Cov.)	Â12 (Cov.)	Median AUC (Baseline)	Median AUC (Exp.)	Adj. p-value (AUC)	Â12 (AUC)	Union Cov. (Baseline)	Union Cov. (Exp.)	Execs/s (Baseline)	Execs/s (Exp.)
ldk	24	5	5	17166	37581	0.015873	0.96	401693	850156	0.015873	1	36760	38118	15.06	13.05
lnd	24	5	5	26946	30011	0.015873	1	642933	717990	0.015873	1	27186	30281	3.67	3.36

A comprehensive version of this table including raw P-values and Interquartile Ranges (IQRs) is available in evaluation_metrics.csv.

2. Interpretation Guide

Use the generated matrix above to objectively evaluate the experimental configuration. For full methodology, see the Smite Fuzzing Evaluation Framework.

Key Metrics

• Adj. p-value: Mann-Whitney U test corrected for multiple targets via Holm-Bonferroni. Controls false-positive rate to ≤ 5% across all targets.
• Â12: Probability that a random B trial outperforms a random A trial. 0.5 = no difference; 0.7 = B wins 70% of pairings. Always read alongside the p-value.
• IQR: Spread of the middle 50% of trials. A much larger IQR in B suggests a few outlier runs may be inflating the median.
• AUC: Coverage speed — how much was discovered and how early. Useful when final coverage is similar between configurations.
• Union Coverage: Union of all trial bitmaps; the coverage ceiling for a multi-core deployment. Descriptive only, cannot be statistically tested.
• Execs/s: A large drop in B without a coverage gain means the new feature is too expensive.

Reading the Results

Adj. p	Â12	Conclusion
< 0.05	> 0.5	Meaningful improvement. Check IQRs are comparable, then merge.
< 0.05	~0.5	Significant but negligible. Check if worth the added complexity.
> 0.05	> 0.6	Promising but underpowered. Re-run with more trials (e.g., 50).
> 0.05	~0.5	No effect. Try an advanced snapshot or ground-truth evaluation.
any	< 0.5	B underperforms A. If significant, reject or redesign the feature.

Time-series caveat: If the IQR bands overlap for most of the campaign and only diverge near the end, treat the final-coverage result cautiously — late divergence may reflect noise rather than a sustained advantage.

3. Visualizations

Note: In the box plots below, the central box represents the Interquartile Range (IQR, the middle 50% of trials), demonstrating the consistency of the fuzzer's performance. The internal line represents the median.

Target: ldk

Median Coverage Over Time

Distribution Comparisons

Final Edge Coverage	Area Under Curve (Speed)

---

Target: lnd

Median Coverage Over Time

Distribution Comparisons

Final Edge Coverage	Area Under Curve (Speed)

---

[erickcestari]

LGTM!

edit: Needs rebase.

[morehouse]

Wow, I just fell off my chair. Double the coverage looks good to me.

[NishantBansal2003]

CLN and Eclair are also on the way, I'll share their results as well in some time