[vorstand-boris-bot]

This makes the market maker's slowest daemon test cheaper, so it stops timing out under CI's load.

Why

a_market_maker_moves_a_skewed_core_node_toward_balance_and_gains (from #730) runs in about 13 s locally. In CI it took 83 s on a run that passed (#771's rerun). It hit nextest's 120 s limit on #771's first run, on #779, and on fix/hub-account-state-after-final, all around 17:25–17:35 UTC on 2026-11-04. The two-target test takes 17 s in CI.

Timing its phases on one core (taskset -c 0) showed where the time goes. Most of it was the market maker swapping about $20 through d1 before its 2-point threshold called for a rebalance. That's hundreds of packets over both hubs.

Change

The threshold goes from 2 points to half a point (5_000 ppm). The market maker then rebalances on the pool after one or two trades. The test still asserts the same things:

• d1's DSUSD share rises;
• the market maker's XRP goes back through the pool;
• the market maker's value grows at the pool's price;
• the books are unchanged across a restart.

On one core, the trade phase falls from 7.5 s to 1.9 s, and the whole test from 15.4 s to 11.0 s. The rest is mostly the harness's claimed and settled waits before and after the restart, which every hub test shares. I left those alone.

Both daemon tests pass. Only the test file changes.

🤖 Generated with Claude Code

[vorstand-edna-bot]

Approve: a test-input change only; the test keeps every assertion, and still exercises a trade through d1 followed by a rebalance on the pool.

Blocking

None.

Non-blocking

1. crates/dassie-cli/tests/market_maker.rs:199: the reworded comment line runs past the file's usual width (rustfmt won't rewrap it); worth reflowing.
2. For the record: thresholdPpm is a config input the test already set below the default (20 000 vs 50 000), not a production default, and 5 000 is a valid value (config.rs refuses only 0 and above MAX_THRESHOLD_PPM), so this doesn't tune a production rule to suit the test. At the maker's ~$1000 starting even, one CAP-sized trade already crosses the 0.5% limit in beyond_threshold, as the description says. Trading several times before a rebalance is still covered by the two-target test, which runs at the default threshold.