The v2.3.0 campaign — plan¶

Status: complete (2026-10-18). One default moves: quiet mode (EnableQuietMode) ships on. The results are at the end of this page. Above them, unchanged, is the pre-registration: it fixed the arms, the cells and the decision rule before the first dispatch, as the v2.2.0 campaign did. The plan file is tools/bench/campaigns/v2.3.0.yaml, and the jobs were dispatched and collected with tools/bench/campaign.py (#662).

← Benchmark index · Methodology · Roadmap

What this campaign decides¶

v2.3.0 adds regime-gated mechanisms. Each ships default off and is byte-identical when off (the ADR-0020 check, run on each mechanism's PR). Each targets a gap a published page measured:

arm	mechanism	issue	target regime	the gap it targets
`ll`	link-lifetime prediction: discount a next hop by its predicted remaining lifetime (`EnableLinkLifetime`, range per cell)	#574	VANET	42.1 % vs the oracle's 81.7 % on VANET main (vanet.md)
`quiet`	quiet mode: back proactive ants off on a stable route (`EnableQuietMode`)	#571	static mesh	NRL 4.51 vs OLSR's 1.78 (static-mesh.md)
`evap`	adaptive evaporation: each destination's decay follows its route churn (`EnableAdaptiveEvaporation`)	#572	LEO shells, VANET	walker16's 2.07 hop changes per flow-minute vs the oracles' ~1.0 (leo-walker.md)
`energy`	the energy-aware metric (`LinkMetric=energy`)	#145	battery-limited fields	not yet a harness metric; see below
`timing`	the propagation timing profile: `T_hop` = mean ISL propagation (`--timingProfile=propagation`)	#205	satellite	15.5 pp under the delay oracle on Starlink S1, 7.0 pp on walker16
`p2phello`	a hello every 5th tick on point-to-point ISLs (`P2pHelloEvery=5`)	#204	static ISL grid	hellos cost NRL and discover nothing on a fixed link (v2.2.0's `hello500` data)

The cells¶

The 18 cells of the v2.2.0 corpus, with their recorded dispatch inputs: - the MANET grid (6); - the static mesh; - FANET main and sparse; - VANET main (five 4-seed blocks) and sparse (two 10-seed blocks); - the four ISL-grid cells; - the three LEO shells.

Every cell runs 20 seeds. The -opt image runs everything except leo-walker, which runs on ns-3.48.

Arms and where they run¶

arm	protocols	cells	why
`control`	every protocol the cell publishes	all 18	the shipped defaults at the campaign commit; also v2.3.0's corpus if no default moves
`ll`, `quiet`, `evap`, `energy`	`anthocnet`	all 18	ADR-0019: a mechanism is measured in every family, not only its target
`timing`	`anthocnet`	4 ISL-grid + 3 LEO	the profile is a satellite-harness option; on Wi-Fi cells T_hop is the thesis's 3 ms
`p2phello`	`anthocnet`	4 ISL-grid	safe only where a failed ISL takes its interface down. leo-walker's failure overlay drops packets on a link that stays up, which a throttled hello would hide (configuration.md, `P2pHelloEvery`)

Notes on what each arm means per cell, stated before any number exists: - ll on the satellite cells: point-to-point ISLs have no radio range. The plan uses 5000 km, about the longest ISL, as a geometric proxy. On the static ISL grid nothing moves, so no link is predicted to end, and the arm should read as a no-op there. On the LEO shells it discounts links whose endpoints are separating. The pages say so next to any satellite ll number. - energy: the Wi-Fi harness gives each node 5000 J, sized so that no node dies in 900 s (anthocnet-compare.cc, kDefaultEnergyJ). Residual energy then stays above ~85 %, so the metric prices hops at about 1.0–1.2, and it behaves close to hop count. The arm measures what the metric costs when batteries are not the binding constraint. A battery-limited cell (lower --energyJ, first-node-death as the outcome) is follow-up work, not this campaign. On the satellite harnesses no energy probe is set, so the metric is plain hop count, as sinr was there in v2.2.0. - quiet on mobile cells: routes change, so the back-off should rarely engage, and the arm should read as noise. A loss there is the finding to look for.

Pre-registered decision rule¶

The rule is v2.2.0's, unchanged: - every comparison is paired per seed against control, on the same 20 seeds, with campaign.py verdicts (bench_parse.py --ab); - a default moves only if no cell is PAIRED-WORSE on PDR, no cell is PAIRED-WORSE on delay99 by more than 10 % while PDR is NOISE, and at least one family is PAIRED-IMPROVED; - otherwise the mechanism ships selectable and off, and its per-family deltas go into the mechanism × regime table in network-regimes.md.

v2.3.0's own exit criterion, from the roadmap, adds one line: each mechanism must show a paired improvement in its target family. A mechanism that does not is recorded on its issue as not earning its place, and the issue stays open with the data.

For quiet and p2phello, whose target is overhead, "improvement" is NRL PAIRED-IMPROVED with PDR not PAIRED-WORSE in the target cells.

Attribution: in one cell per family, bench_parse.py --control compares the control's baseline rows against v2.2.0's corpus. Every baseline row must be identical, except stock AODV on the satellite cells, which may read ORDER-ONLY (#675). Anything else means the harness moved, and that family's deltas are not attributable until it is explained.

If two or more defaults move, a combined arm runs in all 18 cells with every protocol, and that generation is what the pages publish.

Phase order¶

Code, no runs. Merge each mechanism with its core tests and the byte-identity evidence. Merge this plan. Preflight every cell (0 FAIL since #663).
Screen (done before this plan): ll on VANET main and sparse and the two Nakagami grid cells, paired against a same-commit control. Its result is recorded on #574. It does not decide anything; it only checked that the mechanism acts.
Campaign: all 139 jobs at main, at most 12 runs active (#662). Each arm's verdict goes on its issue when its last cell lands (ADR-0013).
Decide and restate: apply the rule, fill the mechanism × regime table, restate the family pages if a default moved, and update the roadmap row.

What would make this plan wrong¶

The control disagrees with the v2.2.0 corpus on a row the #675 control calls a FAIL. Stop: the harness moved.
An arm's attribute does not reach the run (its ##CONFIG## attr line shows the default). The cell is void; re-dispatch it.
ll is identical to control on VANET. The hellos then carry no kinematics, or the range is wrong: diagnose before spending more seeds.

Results¶

Verdict: quiet mode's default moves to on; the other five ship selectable and off. Quiet mode is the only arm with no PAIRED-WORSE cell on any metric, and it is PAIRED-IMPROVED in two families (static mesh, satellite ISL grid), so it meets rule 1. It also meets the exit criterion in its target family: on the static mesh, NRL falls from 4.51 to 4.11 (−8.6 %, paired 95 % CI [−0.47, −0.32]) and delay99 from 10.3 to 9.6 ms, with PDR unchanged. Only one default moved, so the combined arm does not run: the v2.3.0 generation is control for every other protocol and the quiet arm for AntHocNet.

arm	attribute	rule 1	target family	where it helps	issue
`quiet`	`EnableQuietMode=true`	passes: no cell worse on any metric	met: static mesh NRL −0.40 (−8.6 %)	static mesh; ISL grid NRL −0.04 to −0.11	#571
`ll`	`EnableLinkLifetime=true`	fails: Starlink S1 PDR −0.49 pp	not met: VANET main NRL +1.67, PDR noise; sparse noise	MANET grid: Nakagami PDR +2.1 to +3.9 pp, delay99 −159 to −215 ms	#574
`evap`	`EnableAdaptiveEvaporation=true`	fails: grid Gauss-Markov Nakagami −1.12 pp, LEO shells −0.5 to −5.5 pp	not met: LEO and VANET both worse	static mesh NRL −0.14; FANET main PDR +0.41 pp	#572
`energy`	`LinkMetric=energy`	fails: VANET main −1.52 pp	not measured: no battery-limited cell in this campaign	FANET main, grid Gauss-Markov two-ray: about +0.5 pp, at a delay cost	#145
`timing`	`--timingProfile=propagation`	no regression, but no improvement anywhere	not met: noise on all 7 satellite cells	—	#205
`p2phello`	`P2pHelloEvery=5`	fails: sat-seam −0.12 pp	partly met: NRL −0.5 to −10.5 on base, failcell and corridor; seam loses PDR	ISL grid without unscheduled failures	#204

The mechanism × regime summary is section 8 of network-regimes.md.

Validity¶

Commits. The runs ran at main between 54cdfae7 and cc230126 (2026-10-17 to 2026-10-18), at most 12 active. The only change under core/, ns3/ or the benchmark workflows in that window is the ns-3 replay instrumentation (#641): a read-only StrongestNextHop accessor and the harnesses' --vizTrace flag, which schedules nothing unless given. The attribution control below confirms no behaviour moved.
Attribution. bench_parse.py --control against v2.2.0's corpus, one cell per family (grid rwp/tworay and Gauss-Markov/Nakagami, static mesh, FANET main, VANET main, sat-base, walker16 AODV, Starlink S1 oracle): every baseline row is identical on 20 seeds. The exceptions are stock AODV on sat-base and walker16, which read ORDER-ONLY with paired deltas inside noise, as #675 allows.
Every arm reached its run. On the Wi-Fi cells each arm file's ##CONFIG## attr line shows the arm's value. The satellite harnesses print no attribute lines, so there the check is that the arm's rows differ from control's. They do for every arm and cell except ll on the four ISL-grid cells, which are byte-identical to control. The plan predicted exactly that: nothing moves on the torus, so no link is predicted to end.
scenario_check.py results on the 113 cells gives 35 FAIL, all the known #431 oracle-hop findings on the Nakagami cells. v2.2.0's corpus gives the same 35, as it must: those baseline rows are identical.
Byte-identical cells. Besides ll on the ISL grid, quiet is byte-identical to control on the three Nakagami grid cells: no route stays stable for one proactive round there, so the back-off never engages.
The OLSR walker16 control. Both dispatches (runs 39036845475 and 39114244498) were cancelled at the 6 h job limit. It is a baseline row and enters no mechanism verdict. OLSR is since retired as a LEO arm: baselines are family-native (ADR-0026), and the link-state slot on the LEO shells goes to OPSPF (#707).
Pairing. Every comparison is paired per seed against control, n = 20 in every cell. VANET main and sparse ran in seed blocks (5 × 4 and 2 × 10), concatenated by campaign.py collect and paired by seed.

Two findings beyond the verdicts¶

The energy metric at full batteries behaves like sinr. With 5000 J per node, residual energy stays high and every hop costs about 1.0 to 1.2. The arm's losses match v2.2.0's LinkMetric=sinr arm cell for cell in sign and size (VANET main −1.5 pp, delay99 +1 s, NRL +13; Starlink S1 +0.7 pp at NRL +49), though no row is identical. What costs is leaving hop-count goodness, not the energy signal. A battery-limited cell is the measurement #145 still needs.
Link lifetime pays on fades, not on corners. Its largest gains are on the Nakagami grid cells, where it steers off links about to fade. On VANET, its target, the corner break is a topology change that kinematics alone does not predict.

Per arm, per cell¶

Each row is AntHocNet in the arm minus AntHocNet in control, paired per seed. dPDR is in percentage points, d delay99 in ms with a bootstrap interval, and d NRL a t-interval. The mark after each interval reads + when the interval lies above zero, − when it lies below, and · when it straddles zero. The last column is bench_parse.py's own verdict.

`ll` (#574)¶

cell	n	dPDR	d delay99	d NRL	verdict
grid-rwp-tworay	20	+0.29 [+0.06, +0.52] pp +	-27.75 [-30.45, -24.95] ms −	+0.10 [+0.02, +0.19] +	PAIRED-MIXED
grid-rwp-nakagami	20	+2.10 [+1.34, +2.85] pp +	-158.50 [-185.65, -132.20] ms −	-1.05 [-1.30, -0.80] −	PAIRED-IMPROVED
grid-ssrwp-tworay	20	+0.33 [+0.11, +0.56] pp +	-29.55 [-32.35, -26.80] ms −	+0.08 [-0.03, +0.19] ·	PAIRED-IMPROVED
grid-ssrwp-nakagami	20	+2.57 [+1.29, +3.86] pp +	-163.00 [-196.85, -129.75] ms −	-1.21 [-1.59, -0.82] −	PAIRED-IMPROVED
grid-gaussmarkov-tworay	20	+0.73 [+0.45, +1.01] pp +	-36.60 [-40.90, -32.80] ms −	+0.03 [-0.09, +0.15] ·	PAIRED-IMPROVED
grid-gaussmarkov-nakagami	20	+3.91 [+2.71, +5.10] pp +	-214.95 [-265.20, -165.50] ms −	-1.77 [-2.37, -1.16] −	PAIRED-IMPROVED
static-mesh	20	-0.20 [-0.48, +0.08] pp ·	+0.70 [+0.35, +1.15] ms +	+0.08 [-0.03, +0.18] ·	PAIRED-WORSE
fanet-main	20	+0.50 [+0.23, +0.77] pp +	-10.85 [-14.30, -7.50] ms −	-0.02 [-0.05, +0.01] ·	PAIRED-IMPROVED
fanet-sparse	20	+0.16 [-0.15, +0.47] pp ·	+2.25 [-5.90, +10.75] ms ·	-0.00 [-0.07, +0.07] ·	PAIRED-NOISE
vanet-main	20	-0.13 [-0.47, +0.21] pp ·	+25.10 [-3.70, +54.05] ms ·	+1.67 [+0.72, +2.62] +	PAIRED-WORSE
vanet-sparse	20	+0.01 [-0.15, +0.17] pp ·	+13.85 [-26.50, +53.40] ms ·	+0.14 [-0.06, +0.34] ·	PAIRED-NOISE
sat-base	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	+0.00 [+0.00, +0.00] ·	PAIRED-NOISE
sat-failcell	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	+0.00 [+0.00, +0.00] ·	PAIRED-NOISE
sat-corridor	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	+0.00 [+0.00, +0.00] ·	PAIRED-NOISE
sat-seam	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	+0.00 [+0.00, +0.00] ·	PAIRED-NOISE
leo-walker16-anthocnet	20	-0.13 [-0.42, +0.15] pp ·	+0.95 [-13.25, +13.40] ms ·	+0.19 [-1.08, +1.46] ·	PAIRED-NOISE
leo-starlink1-anthocnet	20	-0.49 [-0.88, -0.11] pp −	+3.60 [+0.05, +7.30] ms +	+36.54 [+11.93, +61.15] +	PAIRED-WORSE
leo-starlink1-storm-anthocnet	20	-0.00 [-0.74, +0.74] pp ·	-0.20 [-5.00, +4.50] ms ·	-4.44 [-32.21, +23.33] ·	PAIRED-NOISE

`quiet` (#571)¶

cell	n	dPDR	d delay99	d NRL	verdict
grid-rwp-tworay	20	-0.05 [-0.16, +0.07] pp ·	+0.15 [-2.35, +2.85] ms ·	-0.05 [-0.12, +0.03] ·	PAIRED-NOISE
grid-rwp-nakagami	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	+0.00 [+0.00, +0.00] ·	PAIRED-NOISE
grid-ssrwp-tworay	20	-0.21 [-0.47, +0.06] pp ·	-1.55 [-4.10, +0.95] ms ·	-0.01 [-0.07, +0.05] ·	PAIRED-NOISE
grid-ssrwp-nakagami	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	+0.00 [+0.00, +0.00] ·	PAIRED-NOISE
grid-gaussmarkov-tworay	20	-0.10 [-0.35, +0.15] pp ·	+1.45 [-4.85, +7.75] ms ·	+0.07 [-0.02, +0.16] ·	PAIRED-NOISE
grid-gaussmarkov-nakagami	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	+0.00 [+0.00, +0.00] ·	PAIRED-NOISE
static-mesh	20	-0.00 [-0.25, +0.24] pp ·	-0.70 [-1.00, -0.45] ms −	-0.40 [-0.47, -0.32] −	PAIRED-IMPROVED
fanet-main	20	+0.04 [-0.18, +0.25] pp ·	+0.85 [-2.20, +4.10] ms ·	-0.01 [-0.03, +0.01] ·	PAIRED-NOISE
fanet-sparse	20	-0.14 [-0.38, +0.11] pp ·	-1.55 [-5.60, +3.05] ms ·	+0.02 [-0.05, +0.09] ·	PAIRED-NOISE
vanet-main	20	-0.01 [-0.02, +0.01] pp ·	+1.60 [+0.00, +4.80] ms ·	+0.02 [-0.02, +0.06] ·	PAIRED-NOISE
vanet-sparse	20	+0.03 [-0.07, +0.14] pp ·	+17.15 [-9.70, +43.35] ms ·	-0.03 [-0.16, +0.10] ·	PAIRED-NOISE
sat-base	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	-0.04 [-0.05, -0.04] −	PAIRED-IMPROVED
sat-failcell	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	-0.05 [-0.05, -0.04] −	PAIRED-IMPROVED
sat-corridor	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	-0.11 [-0.12, -0.09] −	PAIRED-IMPROVED
sat-seam	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	-0.11 [-0.12, -0.10] −	PAIRED-IMPROVED
leo-walker16-anthocnet	20	-0.11 [-0.42, +0.20] pp ·	+1.05 [-14.15, +14.25] ms ·	+0.03 [-1.26, +1.33] ·	PAIRED-NOISE
leo-starlink1-anthocnet	20	-0.09 [-0.58, +0.39] pp ·	+3.90 [-2.35, +10.40] ms ·	+18.10 [-13.55, +49.75] ·	PAIRED-NOISE
leo-starlink1-storm-anthocnet	20	+0.38 [-0.59, +1.35] pp ·	-3.30 [-7.95, +2.00] ms ·	-23.24 [-76.58, +30.09] ·	PAIRED-NOISE

`evap` (#572)¶

cell	n	dPDR	d delay99	d NRL	verdict
grid-rwp-tworay	20	-0.01 [-0.20, +0.19] pp ·	+1.20 [-0.65, +3.00] ms ·	-0.04 [-0.11, +0.03] ·	PAIRED-NOISE
grid-rwp-nakagami	20	-0.37 [-1.11, +0.36] pp ·	+50.55 [+18.75, +81.50] ms +	+1.30 [+1.03, +1.58] +	PAIRED-WORSE
grid-ssrwp-tworay	20	-0.10 [-0.30, +0.11] pp ·	+0.75 [-1.50, +2.95] ms ·	+0.03 [-0.05, +0.10] ·	PAIRED-NOISE
grid-ssrwp-nakagami	20	+0.02 [-1.02, +1.07] pp ·	+53.60 [+12.30, +89.80] ms +	+1.29 [+1.04, +1.55] +	PAIRED-WORSE
grid-gaussmarkov-tworay	20	-0.06 [-0.43, +0.30] pp ·	+2.35 [-3.85, +8.95] ms ·	-0.04 [-0.13, +0.05] ·	PAIRED-NOISE
grid-gaussmarkov-nakagami	20	-1.12 [-1.92, -0.33] pp −	+58.65 [+15.45, +101.95] ms +	+1.97 [+1.39, +2.56] +	PAIRED-WORSE
static-mesh	20	+0.05 [-0.37, +0.46] pp ·	-0.10 [-0.35, +0.20] ms ·	-0.14 [-0.21, -0.08] −	PAIRED-IMPROVED
fanet-main	20	+0.41 [+0.12, +0.70] pp +	-0.20 [-4.30, +3.50] ms ·	-0.03 [-0.06, +0.01] ·	PAIRED-IMPROVED
fanet-sparse	20	-0.14 [-0.47, +0.18] pp ·	+13.05 [+5.65, +21.45] ms +	+0.06 [-0.02, +0.14] ·	PAIRED-WORSE
vanet-main	20	-0.02 [-0.25, +0.20] pp ·	+180.05 [+158.60, +200.35] ms +	+1.46 [+0.70, +2.22] +	PAIRED-WORSE
vanet-sparse	20	-0.03 [-0.16, +0.11] pp ·	+108.05 [+73.80, +141.80] ms +	+0.10 [-0.05, +0.25] ·	PAIRED-WORSE
sat-base	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	+0.00 [-0.00, +0.01] ·	PAIRED-NOISE
sat-failcell	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	+0.00 [-0.00, +0.00] ·	PAIRED-NOISE
sat-corridor	20	+0.00 [+0.00, +0.00] pp ·	-4.15 [-12.45, +0.00] ms ·	+0.01 [-0.00, +0.02] ·	PAIRED-NOISE
sat-seam	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	-0.01 [-0.01, -0.00] ·	PAIRED-IMPROVED
leo-walker16-anthocnet	20	-0.48 [-0.89, -0.07] pp −	+20.15 [-2.05, +45.35] ms ·	+2.23 [-1.55, +6.01] ·	PAIRED-WORSE
leo-starlink1-anthocnet	20	-5.40 [-7.07, -3.73] pp −	+18.10 [+10.10, +27.05] ms +	+235.57 [+87.57, +383.58] +	PAIRED-WORSE
leo-starlink1-storm-anthocnet	20	-5.47 [-6.87, -4.08] pp −	+11.00 [+5.35, +16.50] ms +	+164.26 [+74.83, +253.68] +	PAIRED-WORSE

`energy` (#145)¶

cell	n	dPDR	d delay99	d NRL	verdict
grid-rwp-tworay	20	-0.07 [-0.24, +0.11] pp ·	+7.75 [+4.90, +10.95] ms +	+0.21 [+0.15, +0.27] +	PAIRED-WORSE
grid-rwp-nakagami	20	-0.50 [-1.37, +0.37] pp ·	+127.55 [+83.90, +171.45] ms +	+1.35 [+1.01, +1.70] +	PAIRED-WORSE
grid-ssrwp-tworay	20	+0.05 [-0.17, +0.27] pp ·	+9.75 [+6.35, +13.50] ms +	+0.21 [+0.12, +0.30] +	PAIRED-WORSE
grid-ssrwp-nakagami	20	-0.52 [-1.48, +0.44] pp ·	+139.25 [+84.35, +198.40] ms +	+1.55 [+1.09, +2.02] +	PAIRED-WORSE
grid-gaussmarkov-tworay	20	+0.51 [+0.34, +0.68] pp +	+28.20 [+19.85, +36.35] ms +	+0.23 [+0.14, +0.33] +	PAIRED-MIXED
grid-gaussmarkov-nakagami	20	-0.73 [-1.79, +0.33] pp ·	+126.35 [+68.50, +178.30] ms +	+2.29 [+1.51, +3.08] +	PAIRED-WORSE
static-mesh	20	+0.19 [-0.14, +0.52] pp ·	+0.20 [+0.00, +0.40] ms ·	-0.04 [-0.11, +0.03] ·	PAIRED-NOISE
fanet-main	20	+0.48 [+0.21, +0.75] pp +	+19.85 [+14.65, +25.50] ms +	+0.04 [+0.00, +0.07] ·	PAIRED-MIXED
fanet-sparse	20	+0.21 [-0.06, +0.48] pp ·	+564.00 [+499.70, +629.90] ms +	+0.37 [+0.27, +0.47] +	PAIRED-WORSE
vanet-main	20	-1.52 [-1.71, -1.33] pp −	+1029.20 [+998.60, +1061.40] ms +	+12.56 [+11.61, +13.50] +	PAIRED-WORSE
vanet-sparse	20	+0.50 [+0.35, +0.66] pp +	+1553.95 [+1506.50, +1597.70] ms +	+1.42 [+1.20, +1.63] +	PAIRED-MIXED
sat-base	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	+0.00 [-0.00, +0.00] ·	PAIRED-NOISE
sat-failcell	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	+0.00 [-0.00, +0.00] ·	PAIRED-NOISE
sat-corridor	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	+0.00 [+0.00, +0.00] ·	PAIRED-NOISE
sat-seam	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	+0.00 [+0.00, +0.01] ·	PAIRED-WORSE
leo-walker16-anthocnet	20	-0.01 [-0.28, +0.27] pp ·	+0.35 [-10.25, +9.20] ms ·	+1.95 [+0.68, +3.22] +	PAIRED-WORSE
leo-starlink1-anthocnet	20	+0.70 [+0.30, +1.10] pp +	+6.15 [+1.50, +10.75] ms +	+49.44 [+25.32, +73.55] +	PAIRED-MIXED
leo-starlink1-storm-anthocnet	20	+0.41 [-0.43, +1.25] pp ·	+8.55 [+0.80, +16.15] ms +	+36.23 [-9.22, +81.67] ·	PAIRED-WORSE

`timing` (#205)¶

cell	n	dPDR	d delay99	d NRL	verdict
sat-base	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	+0.00 [-0.00, +0.00] ·	PAIRED-NOISE
sat-failcell	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	+0.00 [-0.00, +0.00] ·	PAIRED-NOISE
sat-corridor	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	+0.00 [+0.00, +0.00] ·	PAIRED-NOISE
sat-seam	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	+0.00 [-0.00, +0.00] ·	PAIRED-NOISE
leo-walker16-anthocnet	20	-0.06 [-0.28, +0.16] pp ·	-4.70 [-16.60, +4.75] ms ·	-0.12 [-1.09, +0.86] ·	PAIRED-NOISE
leo-starlink1-anthocnet	20	-0.17 [-0.60, +0.26] pp ·	+2.10 [-2.85, +6.90] ms ·	+5.84 [-13.78, +25.47] ·	PAIRED-NOISE
leo-starlink1-storm-anthocnet	20	-0.25 [-1.08, +0.58] pp ·	+3.35 [-1.50, +8.40] ms ·	+0.02 [-57.91, +57.95] ·	PAIRED-NOISE

`p2phello` (#204)¶

cell	n	dPDR	d delay99	d NRL	verdict
sat-base	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	-0.49 [-0.56, -0.42] −	PAIRED-IMPROVED
sat-failcell	20	+0.00 [+0.00, +0.00] pp ·	+0.00 [+0.00, +0.00] ms ·	-0.46 [-0.53, -0.40] −	PAIRED-IMPROVED
sat-corridor	20	+0.00 [+0.00, +0.00] pp ·	+3.50 [-12.10, +19.10] ms ·	-10.45 [-11.52, -9.39] −	PAIRED-IMPROVED
sat-seam	20	-0.12 [-0.22, -0.01] pp −	+9.00 [+6.50, +11.50] ms +	-0.18 [-0.51, +0.16] ·	PAIRED-WORSE