This table summarizes protocol traps that change client decisions. Observed behavior is limited to its dated scope; documentation-only statuses and toolkit safeguards remain labeled separately.
TL;DR
Keep exact IDs, original sent quantity, complete page evidence, signed positions, and unknown outcomes. No client-ID retry rule can substitute for resolution.
1. Dated findings
| Date / evidence | Trap | Rule | Detail |
|---|---|---|---|
| Observed 2026-10-01 | Repeated client ID creates another order | Never replay an unknown placement | orders.md §2 |
| Observed 2026-10-01 | Filled event quantity is one execution | Sent minus remaining; deduplicated trades | orders.md §3 |
| Observed 2026-10-01 | FILLED can retain a remainder | Await terminal evidence | orders.md §3 |
| Observed 2026-10-01 | NO_SUCH_ORDER has zero/default fields | It proves neither fill nor clean cancel | orders.md §4 |
| Observed 2026-10-01 | Resting reduce-only survives flat exposure | Do not assume its displayed quantity is capped | orders.md §5 |
| Observed 2026-10-01 | Capacity combines both sides | 20 resting orders per symbol total | orders.md §5 |
| Observed 2026-10-01 | LIMIT GTC waits for speed bump | Passive LIMIT is delayed too | orders.md §1 |
| Observed 2026-10-01 | Crossing ALO cancels after ACK | ACK does not mean resting | orders.md §1 |
| Observed 2026-10-01 | STP removes resting self-order | Incoming IoC may continue matching | orders.md §4 |
| Observed 2026-10-01 | Historic order appears late | History miss is not rejection | orders.md §6 |
| Observed 2026-10-01 | REST order counts lag socket | Temporary mismatch is not wrong-account proof | account-and-leverage.md §2 |
| Observed 2026-10-01 | Position rows are signed and can remain zero | Do not infer size from a separate side | account-and-leverage.md §1 |
| Observed 2026-10-01 | Leverage fails while symbol occupied | Lock by symbol, not account | account-and-leverage.md §4 |
| Observed 2026-10-01 | UUID case affects selection | Use canonical server account ID | api-and-auth.md §2 |
| Observed 2026-10-01 | Auth result differs from one docs example | Accept known explicit variants | api-and-auth.md §3 |
| Observed 2026-10-01 | Numeric strings/nonboolean flags rejected | Serialize numeric literals and boolean flags | orders.md §1 |
| Public probe 2026-09-30 | Refdata cached beyond its immediate fetch | Live band for IoC clamp | markets-and-numbers.md §4 |
| Public probe 2026-09-30 | clobPairId repeats | Full symbol identity | markets-and-numbers.md §1 |
| Public probe 2026-09-30 | REST book padded with zero levels | Validate filtered bid/ask pair | market-data.md §3 |
| Public probe 2026-10-03 | Inventory changes | Query current listings | markets-and-numbers.md §1 |
| Documentation | Broad cancellation affects account | Separate explicit scope from exact cancel | orders.md §4 |
| Implementation inference | Read types have no request ID | One outstanding request per response kind | api-and-auth.md §3 |
| Experimental toolkit | Fill journal has limited freshness evidence | Opt-in, uncertainty exposed, never proof | orders.md §6 |
2. Diagnostic families
Distinguish InvalidJSONFormat wire errors, broad InvalidOrder, tick/quantity bounds, FAILED_MARGIN_CHECK, symbol risk/open-interest limits, order-capacity rejection, RateLimited, and KYC/terms/permission conditions. Unknown names remain visible. A server error after ACK cannot erase acceptance evidence.
Documented and observed frame/reply variants can differ. Preserve compatible decoding rather than assume one page is exhaustive. The fake exchange's adverse controls stress the client and must not become dated venue claims.
Open questions / not verified
- Resting reduce-only matching at zero exposure, subaccount scope, throttle shape/limits, stop visibility, equity freshness/floor, and stream cadence.
- Optional fill-ledger and binding-probe behavior remains experimental.
Sources
The topic files linked above contain primary URLs and evidence dates. Official index: Official documentation index. Documentation checked through 2026-10-03; venue behavior and response shapes can change.
© markpaper authors. Licensed under CC BY 4.0: when publishing or adapting this material, credit “markpaper — QFEX knowledge base” and link to the original and the license.