Market Data and Order Execution

1.A feed delivers sequence numbers 318, 319, and 321. Each number belongs to a state-changing update on the same channel. What should the book builder infer?

2.You recover a book from a snapshot labeled 440 and have buffered deltas 438 through 445. The snapshot includes all updates through 440. What should you replay?

3.Two redundant multicast lines carry identical updates with the same sequence numbers. Line B delivers 601 after Line A's 601 was applied. What should the merger do?

4.A feed sends incremental deltas where a size of zero removes the specified price level. The best ask at 15.02 becomes zero and the next ask is 15.05. What is the new best ask?

5.A depth feed sends absolute quantities. A level has 700 shares and receives an update setting quantity to 120. What should the book show?

6.At the same price, a market-by-price feed reports aggregate size falling from 900 to 600. What information is missing for an exact queue-position update?

7.A gap affects one feed partition carrying 40 symbols. The other partitions have separate sequences. Which books should be considered potentially incomplete?

8.A snapshot arrives and buffered deltas start one sequence after its boundary, but the buffer overflows during catch-up. What should the service do?

9.The market data socket receives heartbeats continuously, but a specific instrument has stopped receiving expected updates. Which check addresses this failure?

10.Why keep tick prices as scaled integers when rebuilding an order book?

11.A venue resets a feed sequence number to 1 at the start of a new session. Yesterday's last applied number was 74,950. What should deduplication key on?

12.A multicast packet has a starting message sequence of 900 and contains four consecutive messages. The next packet starts at 905. Which message is missing?

13.A parser encounters a message type defined as ignorable by the feed protocol. It still occupies one sequence slot. How should the parser handle it?

14.A feed handler currently publishes a new bid and ask through two independent writes. A reader sometimes sees one side from the previous update. What addresses the consistency problem?

15.A venue sends nanosecond exchange timestamps, but several messages have the same timestamp and distinct sequences. What determines replay order within the channel?

16.Your normalization layer joins market data across venues. One raw symbol maps to different products on two venues. Which identifier avoids mixing them?

17.A same-price bid doubles in size after your update. The feed sends add-order messages, but your book also consumes snapshots as additive updates. What error is most likely?

18.A reconciliation run finds the top of book matches the venue reference but several deeper levels differ. What conclusion is supported?

19.A 500-share order fills 180 shares, then you send a cancel and receive another 70-share fill before cancellation is confirmed. What quantity was canceled if the confirmation cancels all leaves?

20.A cancel request is rejected because an order is already fully filled. You have received only a partial fill locally. What should the gateway do?

21.You submit a new order and the connection drops before an acknowledgement. Why is resending it with a new client order ID risky?

22.An execution report is delivered through the order session and the venue's independent drop copy. What should position accounting use to avoid double counting?

23.An order update reports cumulative executed quantity moving from 35 to 90. The venue defines that field as lifetime filled quantity. How much incremental fill does this imply?

24.A venue confirms a replace of a 100-share order after 40 already filled. The protocol defines the replace quantity as new total order quantity. You request 70. What leaves quantity should you expect?

25.A feed publishes a trade print with price and size but no participant or order IDs. Can your order manager book that as a fill for its order at the same price?

26.A client order ID can be reused by another gateway session. Which lookup key prevents fills from being routed to the wrong strategy?

27.The client requests a cancel while a replace is pending. The protocol requires targeting the currently acknowledged order ID. Which state should the gateway consult?

28.An exchange sends a trade bust that explicitly reverses a previously booked buy of 100 shares. What should accounting do?

29.A buy order is accepted on one venue, while a linked hedge is rejected on another. What should the router assume?

30.The last order state sent to a trader is pending cancel. Which quantity belongs in worst-case exposure until confirmation?

31.An order acknowledgement is processed after a full fill because two local threads delivered callbacks out of order. What should the state machine avoid?

32.A client sees a timeout after sending an order. The venue offers an order status query keyed by the original request identity. What is the useful next step?

33.An execution correction reports the same trade reference with a changed quantity. Why is it insufficient to drop every repeated trade reference as a duplicate?

34.A cancel-on-disconnect setting is enabled. After a connection loss, what should the system do before trusting that all its orders are gone?

35.Your venue supports atomic replace and guarantees the old order stays live if a replace is rejected. What should the gateway retain after such a rejection?

36.An IOC order has expired, and a duplicated partial-fill report arrives after the terminal update. What should happen?

37.An engineer sees stable median tick-to-order latency, but p99 doubles after a new release. What investigation best targets the regression?

38.Your local receive and send timestamps come from a monotonic clock on the same host. The exchange timestamp comes from another clock. Which latency can you compute without synchronizing clocks?

39.One service subtracts exchange event time from its local wall time and occasionally gets negative delays. What should you verify first?

40.During bursts, packets arrive at normal network latency but spend 8 ms waiting before decoding. Which signal best verifies a backlog?

41.A team replaces per-message writes with batches of 32 and sees higher throughput but slower responses when traffic is light. What explains the result?

42.A packet capture shows the gateway sends an order promptly but the acknowledgement arrives late. Which measurement narrows the cause further?

43.A replay benchmark reports one million updates per second after loading all data into memory. What workload is needed to assess burst latency in production?

44.Why might a benchmark understate p99 when its generator waits for each request to finish before sending the next?

45.Latency spikes align with CPU migrations on the market data thread. Which experiment tests the hypothesis most directly?

46.An event consumer misses a latency budget only when the producer and consumer counters share one cache line. What mechanism fits?

47.A service logs every decoded market update synchronously. During volatile periods its decode latency rises when disk writes slow down. What change directly addresses the coupling?

48.A dashboard takes the average of one-minute p99 latency values and labels it daily p99. What should it compute instead?

49.A timer uses wall-clock subtraction across an adjustment by the time synchronization service. The resulting duration is too small. Which clock is appropriate for elapsed time?

50.A latency change saves 10 microseconds in decoding but adds 18 microseconds in the publish stage for the same trace. What is the combined change if other stages stay fixed?

51.A network interface timestamp marks packet arrival before application scheduling. What does subtracting it from an aligned application receive timestamp measure?

52.An allocation-heavy decoder runs fast in short tests but shows periodic long pauses in production. What should you correlate with the spikes?

53.A strategy is long 80 shares with a 120-share limit and has 30 shares of acknowledged buy orders working. The gateway considers the worst-case long position. How many additional buy shares fit?

54.Two gateways share a 100-share remaining limit. Both read 100 remaining and accept 70-share orders. What prevents this race?

55.A kill switch blocks new orders but does not reach orders that were already accepted. What separate action addresses live orders?

56.Your gateway can cancel orders while it lacks fresh data to validate prices. Which action is appropriate during the stale-data condition?

57.A risk check passes using a fresh price, but the order spends 300 ms in a queue and its price reference expires. Where should the gateway check freshness again?

58.A sequence reset clears resend counters but open orders are retained by the venue. What should a restarted gateway restore separately?

59.A backup gateway takes over an order session. The original gateway may still be able to transmit after a network partition. Which control prevents both from sending?

60.A gateway reads its durable journal after restart. The latest order request was logged, but the journal contains no evidence of an acknowledgement. What state should recovery assign until it checks with the venue?

61.A connection's venue sequence is ahead of the last durably processed inbound event. The venue can replay missing messages. Where should recovery begin?

62.A gateway restarts with open-order state but no record of which fills were applied to positions. Which persistence design avoids reapplying or skipping fills?

63.A system uses fixed-width integers for price in ticks and shares for size. Very large values can overflow when multiplied for a notional limit. What should the check do?

64.An order gateway sees valid sizes for most symbols but rejects only one stock after its trading increment changes. Where should you look first?

65.A hedge order is outside the standard price collar during an incident. How should an authorized emergency override be implemented?

66.A gateway is throttled by the venue and has a queue of routine new orders and emergency cancels. The protocol permits both and has one shared message budget. How can the gateway preserve cancel capacity?

67.A per-order limit blocks any order above $500,000. A strategy can send ten $450,000 orders at once. What additional control limits the combined exposure?

68.A strategy gets position updates after fills, while risk checks happen in a separate service. How should accepted orders affect available exposure before fills arrive?

69.A replay is expected to produce the same orders on every run. Identical input packets sometimes produce different decisions when two streams have equal timestamps. What should the harness define?

70.You need to check a change that reconciles venue orders after reconnecting. Which test is most representative?

71.Which property should hold if the same ordinary fill is replayed twice into an order ledger?

72.A strategy consumes a snapshot and later asks for another copy by reference. During a code change, an updater begins mutating a previously published object in place. Which test would catch this change?

73.What is the safest first rollout for a new book builder when the old one already drives live quoting?

74.A book builder falls behind and a bounded queue fills. The queue carries incremental depth changes. Which policy preserves correctness?

75.You discover position mismatches only for partial fills followed by trade busts. What log fields help reconstruct the issue?

76.After a deploy, exchange rejects spike for one release cohort. What comparison best localizes whether the new encoder is involved?

77.A trader reports an order twice in the dashboard, but the venue shows one live order and one execution. Where should you investigate first?

78.You add a circuit breaker for slow risk checks. Which failure injection validates that it protects orders during dependency timeouts?

79.A quote replay diverges at event 10,000. Which debugging method best narrows the root cause?

80.Your counters report 5,000 inbound messages but only 4,970 decoded. Some message types are intentionally ignored. What metric clarifies whether 30 messages were lost?

81.A metrics system labels every latency series with client order ID. The monitoring backend becomes overloaded. What alternative keeps useful latency reporting?

82.A post-incident replay includes trader input and market data but omits exchange acknowledgements and cancel rejects. Why can it miss the bug?

83.A rollout adds a new optional field to persisted snapshots. A rollback may read snapshots written by the new version. What test protects recovery?

84.An order manager restart test waits an arbitrary five seconds for state to settle. It sometimes passes before reconciliation completes. What should the test wait for?

85.A smart router sends a parent buy for 500 shares. One child has filled 150 and another has 200 leaves live. The router wants to avoid overfills even if the second child fills fully. How many shares can it safely allocate elsewhere?

86.A limit price arrives in dollars as 101.25, while a wire field expects hundredths of a dollar. What should the encoder send?

87.A trading system sees a venue quote in cents and another in dollars. Where should it normalize them for route selection?

88.A dashboard marks market data as healthy whenever its queue is empty. The feed process dies and the queue eventually empties. Which additional check detects that state?

89.Two feed adapters provide prices with different timestamp meanings. One stamps matching-engine time and one uses publication time. How should a latency dashboard handle this?

90.Your gateway handles orders for two clients. A resend after reconnection contains reports for both. Which attribute should route each report to a client?

91.You have an order entry feed and a separate drop copy feed. Their respective sequence numbers are 50 and 7,000 for the same execution. Which numbers determine whether those reports duplicate a fill?

92.A venue returns a rate-limit reject for a new order before assigning any venue order ID. What information is needed before scheduling a retry?

93.A trader wants to compare order arrival latency across co-located hosts. Each host has a monotonic clock with a different origin. Which comparison is valid without clock synchronization?

94.A gateway uses multiple threads to update one order. A fill reduces leaves to zero just as a replace acknowledgement writes a nonzero leaves value. What design best prevents this inconsistent outcome?

95.A trading platform stores reference data updates by effective time. Why use the historical effective version in a market replay?

96.A pre-trade check requires a reference price, and a cache returns a price without indicating its source time. What additional data does the check need?

97.A capture confirms that data reached the NIC, yet the feed decoder reports sequence gaps during bursts. Which local measurements help determine whether the host dropped packets?

98.A gateway receives a cancel acknowledgement for an unknown order immediately after startup. What is the appropriate reconciliation response?

99.A process buffers a market event with the assumption it will apply before any subsequent trading decision. How can the decision prove which data it actually used?

100.A failover test verifies that the backup process starts. Which additional outcome establishes trading continuity?

101.A strategy checks that its view of the book is unchanged between making a decision and submitting an order. What metadata supports that check?

102.An order router retries a client request after its HTTP response is lost. The original request may already have created a venue order. What API behavior supports a safe retry?

103.A gateway monitors round-trip time by timing only requests that received acknowledgements. Several requests time out during an incident. What should its report include alongside successful latencies?

104.A market data listener marks an instrument halted, but the order gateway only checks this state at startup. Which change closes the gap?