Session Summary — S45 (session_id=48)¶
Date: 2026-04-21 02:28–02:31 UTC
Duration: 162.94s elapsed (3600s requested)
Halt reason: recovery_exhausted_halt
Phase 7.4 clean session count: 0/2 (halt ≠ duration_elapsed — does not count)
Vitals¶
| Metric | Value |
|---|---|
| Session ID | 48 |
| Ticks | 27 |
| Fills | 2 (buy=1, sell=1) |
| Toxic fills | 0/2 (0.0%) |
| VW realized spread | 10.57 bps |
| Avg realized spread | 10.12 bps |
| Capital start (RLUSD) | 196.979 |
| Capital end (RLUSD) | 196.942 |
| Capital change | −0.037 RLUSD (flat) |
Anchor Regime¶
| Metric | Value |
|---|---|
| Mean error | +9.47 bps |
| Median error | +10.00 bps |
| Range | [+8.1, +10.0] bps |
| Bias | Positive |
| |err| > 5 bps | 100% of ticks |
Verdict: Fully hostile. Same persistent CLOB-AMM divergence as S43 and S44. Regime has been hostile across all three post-FLAG-042 sessions.
Guard Activity¶
Anchor saturation guard fired on entry → engine entered DEGRADED → recovery evaluator cleared the condition → anchor guard fired again on second entry → per-episode cap exhausted → recovery_exhausted_halt.
This is FLAG-042 working as designed. In a fully hostile regime (anchor 100% above 5 bps threshold), a single recovery window cannot sustain — guard re-fires is correct and expected.
FLAG-043 (corridor stability window): Only 27 ticks in session, anchor dominant signal. Corridor contribution to exhaust cycle inconclusive — regime cleared the session before corridor could be isolated. Continue monitoring in S46.
Fills¶
Both fills occurred at the same timestamp (02:30:54 UTC) — simultaneous cross on the same tick:
| Side | Price | Qty |
|---|---|---|
| BUY | 1.4247 | 10.5 XRP |
| SELL | 1.4276 | 19.5 XRP |
Spread captured on the cross: ~20 bps gross. VW 10.57 bps after skew adjustment.
Inventory at Close¶
| Asset | Qty |
|---|---|
| XRP | 64.503 |
| RLUSD | 104.963 |
| Total value (RLUSD) | 196.942 |
| XRP share | 46.7% |
| RLUSD share | 53.3% |
Momentum Filter¶
- Buy suppressions: 2
- Sell suppressions: 0
Pre-S46 Requirements¶
- Realign inventory: Run
python tools/realign_inventory_to_onchain.py --config config/config_live_stage1.yaml --db neo_live_stage1.db --dry-run→ verify zero delta →--confirmif needed. - Regime check: Anchor has been above 8 bps for all of S43–S45. Consider waiting for a better regime window before S46, or accept that guards will fire again.
Notes for Vesper¶
- First native NEO Desk artifact. System is live.
- No engineering action required from this session — engine behavior was correct throughout.
- The persistent hostile regime (S43/S44/S45 all fully above threshold) may warrant an Atlas observation: is CLOB-AMM divergence structural at this time of day, or is this a temporary market condition? Not an escalation yet — flagging for awareness.
— Katja