fix: Do not send off-path data when closing

[flub]

Description

When we are closing we're only supposed to send packets containing
CONNECTION_CLOSE and (PATH_)ACK frames. There is no need to continue
sending any off-path data.

We do have the option of sending a CONNECTION_CLOSE to the new
remote if we get an involuntary migration during close. But we'd
have to make sure not to exceed the anti-amplification limit
since we can no longer validate the path.

We are also allowed to simply discard packets from a migrated
peer during the closing state. Which is what we do.

Breaking Changes

n/a

Notes & open questions

Closes #357

Note that the base is #338. I won't merge before that is merged so
it shows up as a separate commit because this was a pre-existing
bug.

[flub]

Oops, accidentally wrote a monologue in the notes section. Changed back to draft to do the changes.

[github-actions]

Documentation for this PR has been generated and is available at: https://n0-computer.github.io/quinn/pr/358/docs/iroh_quinn/

Last updated: 2026-01-26T15:58:36Z

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 76.77%. Comparing base (c38be74) to head (a14cfac).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #358      +/-   ##
==========================================
- Coverage   76.77%   76.77%   -0.01%     
==========================================
  Files          81       81              
  Lines       22944    22945       +1     
==========================================
  Hits        17615    17615              
- Misses       5329     5330       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions]

Performance Comparison Report

16f81f9534b36617e82f353d75096b218321cb1f - artifacts

Raw Benchmarks (localhost)

Scenario	iroh-quinn	upstream	Delta	CPU (avg/max)
large-single	N/A	8137.2 Mbps	N/A	N/A
medium-concurrent	N/A	7244.4 Mbps	N/A	N/A
medium-single	N/A	4113.6 Mbps	N/A	N/A
small-concurrent	N/A	4932.2 Mbps	N/A	N/A
small-single	N/A	4161.6 Mbps	N/A	N/A

---

6bdbd037c406b3d5a94c869f2ce8b63dee62523a - artifacts

Raw Benchmarks (localhost)

Scenario	iroh-quinn	upstream	Delta	CPU (avg/max)
large-single	5284.6 Mbps	8075.0 Mbps	-34.6%	95.9% / 117.0%
medium-concurrent	5452.1 Mbps	7340.2 Mbps	-25.7%	95.8% / 114.0%
medium-single	3882.2 Mbps	3985.3 Mbps	-2.6%	88.0% / 96.6%
small-concurrent	3864.3 Mbps	5028.2 Mbps	-23.1%	93.0% / 101.0%
small-single	3481.0 Mbps	4223.5 Mbps	-17.6%	87.5% / 96.5%

Netsim Benchmarks (network simulation)

Condition	iroh-quinn	upstream	Delta
ideal	2686.4 Mbps	3491.7 Mbps	-23.1%
lan	768.5 Mbps	796.4 Mbps	-3.5%
lossy	69.8 Mbps	55.9 Mbps	+25.0%
wan	83.8 Mbps	83.8 Mbps	~0%

Summary

iroh-quinn is 22.7% slower on average

---

d342bd05b6995fdc6040afa0aa7fa716e7c0ce1c - artifacts

Raw Benchmarks (localhost)

Scenario	iroh-quinn	upstream	Delta	CPU (avg/max)
large-single	5306.8 Mbps	7699.9 Mbps	-31.1%	98.1% / 182.0%
medium-concurrent	5414.2 Mbps	7178.0 Mbps	-24.6%	92.7% / 99.0%
medium-single	3709.2 Mbps	4130.2 Mbps	-10.2%	93.2% / 123.0%
small-concurrent	3735.4 Mbps	5019.9 Mbps	-25.6%	97.7% / 128.0%
small-single	3276.6 Mbps	4288.4 Mbps	-23.6%	89.1% / 97.0%

Netsim Benchmarks (network simulation)

Condition	iroh-quinn	upstream	Delta
ideal	3016.7 Mbps	3826.9 Mbps	-21.2%
lan	782.4 Mbps	806.3 Mbps	-3.0%
lossy	69.9 Mbps	69.4 Mbps	~0%
wan	83.8 Mbps	83.8 Mbps	~0%

Summary

iroh-quinn is 23.3% slower on average

---

80af1506d01974b41174d788113ea2fa70a80ac2 - artifacts

Raw Benchmarks (localhost)

Scenario	iroh-quinn	upstream	Delta	CPU (avg/max)
large-single	5349.1 Mbps	8150.8 Mbps	-34.4%	92.4% / 97.1%
medium-concurrent	5454.9 Mbps	7922.5 Mbps	-31.1%	96.5% / 118.0%
medium-single	4269.6 Mbps	4577.7 Mbps	-6.7%	96.2% / 116.0%
small-concurrent	3881.8 Mbps	5128.5 Mbps	-24.3%	94.5% / 119.0%
small-single	3414.6 Mbps	4774.6 Mbps	-28.5%	92.9% / 120.0%

Netsim Benchmarks (network simulation)

Condition	iroh-quinn	upstream	Delta
ideal	2849.0 Mbps	3631.6 Mbps	-21.5%
lan	771.4 Mbps	796.4 Mbps	-3.1%
lossy	69.8 Mbps	55.9 Mbps	+25.0%
wan	83.8 Mbps	83.8 Mbps	~0%

Summary

iroh-quinn is 25.6% slower on average

---

a14cfaca6180ff09946ad96c8df6d61b3b04d8ba - artifacts

Raw Benchmarks (localhost)

Scenario	iroh-quinn	upstream	Delta	CPU (avg/max)
large-single	5300.7 Mbps	7676.1 Mbps	-30.9%	91.0% / 96.8%
medium-concurrent	5446.7 Mbps	7623.9 Mbps	-28.6%	95.2% / 109.0%
medium-single	3911.4 Mbps	4463.2 Mbps	-12.4%	93.8% / 109.0%
small-concurrent	3715.8 Mbps	5039.6 Mbps	-26.3%	96.7% / 112.0%
small-single	3276.2 Mbps	4407.5 Mbps	-25.7%	93.2% / 110.0%

Netsim Benchmarks (network simulation)

Condition	iroh-quinn	upstream	Delta
ideal	2890.9 Mbps	3966.3 Mbps	-27.1%
lan	782.5 Mbps	810.4 Mbps	-3.4%
lossy	69.8 Mbps	69.8 Mbps	~0%
wan	83.8 Mbps	83.8 Mbps	~0%

Summary

iroh-quinn is 25.4% slower on average

[divagant-martian]

Has side effects!