A reports an inadequate or failed acceptance condition: Replication task throughput saturates the 1-Gbps corporate link and user latency spikes. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that production network latency stays within limits while the migration backlog decreases predictably. The underlying design requirement is migration bandwidth must be planned around live traffic rather than assuming unlimited private connectivity.

B reports an inadequate or failed acceptance condition: The initial backlog estimate looked acceptable but traffic controls are absent during the business window. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that production network latency stays within limits while the migration backlog decreases predictably. The underlying design requirement is migration bandwidth must be planned around live traffic rather than assuming unlimited private connectivity.

C reports an inadequate or failed acceptance condition: The migration backlog grows because unmonitored replication consumes shared capacity unpredictably. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that production network latency stays within limits while the migration backlog decreases predictably. The underlying design requirement is migration bandwidth must be planned around live traffic rather than assuming unlimited private connectivity.

B would move the database to another subnet in the same AZ and call it independent high availability. This is not a defensible answer to the stated constraint (A rehosted three-tier application runs in one AZ and must satisfy new resilience requirements): modern target architecture should correct inherited single-site risks instead of simply reproducing them on EC2. The required design is redesign with multi-AZ app tiers behind an ALB and a suitable managed multi-AZ data tier; acceptance depends on evidence that aZ-isolation tests show continued checkout success and bounded database recovery.

C would increase the one-AZ VM size rather than distributing compute and data failure domains. This is not a defensible answer to the stated constraint (A rehosted three-tier application runs in one AZ and must satisfy new resilience requirements): modern target architecture should correct inherited single-site risks instead of simply reproducing them on EC2. The required design is redesign with multi-AZ app tiers behind an ALB and a suitable managed multi-AZ data tier; acceptance depends on evidence that aZ-isolation tests show continued checkout success and bounded database recovery.

D would add a redundant ALB but keep web and database resources in a single AZ. This is not a defensible answer to the stated constraint (A rehosted three-tier application runs in one AZ and must satisfy new resilience requirements): modern target architecture should correct inherited single-site risks instead of simply reproducing them on EC2. The required design is redesign with multi-AZ app tiers behind an ALB and a suitable managed multi-AZ data tier; acceptance depends on evidence that aZ-isolation tests show continued checkout success and bounded database recovery.

A reports an inadequate or failed acceptance condition: The web tier survives the AZ drill but the one-AZ database remains unavailable. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that aZ-isolation tests show continued checkout success and bounded database recovery. The underlying design requirement is modern target architecture should correct inherited single-site risks instead of simply reproducing them on EC2.

C reports an inadequate or failed acceptance condition: A larger EC2 instance completes normal requests yet fails completely during zonal isolation. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that aZ-isolation tests show continued checkout success and bounded database recovery. The underlying design requirement is modern target architecture should correct inherited single-site risks instead of simply reproducing them on EC2.

D reports an inadequate or failed acceptance condition: The ALB is multi-AZ but all web servers and the database remain in a single failure zone. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that aZ-isolation tests show continued checkout success and bounded database recovery. The underlying design requirement is modern target architecture should correct inherited single-site risks instead of simply reproducing them on EC2.

A would pick an incompatible RDS edition or licensing option solely for minimum monthly price. This is not a defensible answer to the stated constraint (An on-premises SQL Server service requires minimal administrative overhead but retains SQL compatibility constraints): managed RDS reduces database administration when specialized unsupported engine features are not required. The required design is evaluate RDS for SQL Server support, feature requirements, licensing and migration compatibility; acceptance depends on evidence that a feature-by-feature assessment and representative transaction tests succeed on the managed target.

B would assume RDS for SQL Server supports every on-premises SQL Server feature without assessment. This is not a defensible answer to the stated constraint (An on-premises SQL Server service requires minimal administrative overhead but retains SQL compatibility constraints): managed RDS reduces database administration when specialized unsupported engine features are not required. The required design is evaluate RDS for SQL Server support, feature requirements, licensing and migration compatibility; acceptance depends on evidence that a feature-by-feature assessment and representative transaction tests succeed on the managed target.

D would rehost on a self-managed EC2 SQL Server by default despite the stated operations-reduction goal. This is not a defensible answer to the stated constraint (An on-premises SQL Server service requires minimal administrative overhead but retains SQL compatibility constraints): managed RDS reduces database administration when specialized unsupported engine features are not required. The required design is evaluate RDS for SQL Server support, feature requirements, licensing and migration compatibility; acceptance depends on evidence that a feature-by-feature assessment and representative transaction tests succeed on the managed target.

A reports an inadequate or failed acceptance condition: The cheapest chosen managed edition fails licensing or compatibility acceptance checks. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that a feature-by-feature assessment and representative transaction tests succeed on the managed target. The underlying design requirement is managed RDS reduces database administration when specialized unsupported engine features are not required.

B reports an inadequate or failed acceptance condition: The RDS target lacks a required SQL Server feature used by production transactions. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that a feature-by-feature assessment and representative transaction tests succeed on the managed target. The underlying design requirement is managed RDS reduces database administration when specialized unsupported engine features are not required.

C reports an inadequate or failed acceptance condition: The company continues maintaining a self-managed SQL Server VM, contrary to reduced-operations intent. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that a feature-by-feature assessment and representative transaction tests succeed on the managed target. The underlying design requirement is managed RDS reduces database administration when specialized unsupported engine features are not required.

B would enable ALB round robin while leaving uploads on each server's local filesystem. This is not a defensible answer to the stated constraint (An old application stores user uploads on each application server's local disk, blocking load balancing): shared durable storage removes instance-affinity assumptions and supports horizontal app scaling. The required design is externalize durable files to S3 or suitable shared storage and redesign references accordingly; acceptance depends on evidence that requests hitting different application nodes retrieve the same recently uploaded file.

C would copy new uploads manually between app nodes only during monthly maintenance. This is not a defensible answer to the stated constraint (An old application stores user uploads on each application server's local disk, blocking load balancing): shared durable storage removes instance-affinity assumptions and supports horizontal app scaling. The required design is externalize durable files to S3 or suitable shared storage and redesign references accordingly; acceptance depends on evidence that requests hitting different application nodes retrieve the same recently uploaded file.

D would make every client return to one fixed node instead of providing shared durable files. This is not a defensible answer to the stated constraint (An old application stores user uploads on each application server's local disk, blocking load balancing): shared durable storage removes instance-affinity assumptions and supports horizontal app scaling. The required design is externalize durable files to S3 or suitable shared storage and redesign references accordingly; acceptance depends on evidence that requests hitting different application nodes retrieve the same recently uploaded file.

A reports an inadequate or failed acceptance condition: Upload works on node A but download from node B returns 404. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that requests hitting different application nodes retrieve the same recently uploaded file. The underlying design requirement is shared durable storage removes instance-affinity assumptions and supports horizontal app scaling.

C reports an inadequate or failed acceptance condition: Manual monthly file synchronization leaves recent uploads invisible on other instances. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that requests hitting different application nodes retrieve the same recently uploaded file. The underlying design requirement is shared durable storage removes instance-affinity assumptions and supports horizontal app scaling.

D reports an inadequate or failed acceptance condition: Sticky sessions route a test client to the old node, concealing missing shared file storage. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that requests hitting different application nodes retrieve the same recently uploaded file. The underlying design requirement is shared durable storage removes instance-affinity assumptions and supports horizontal app scaling.

A would send requests to non-durable memory queues and acknowledge before persistence. This is not a defensible answer to the stated constraint (A monolithic queue processor couples request acknowledgement to a slow backend and causes frequent timeouts): asynchronous queuing isolates throughput and availability of upstream and downstream tiers. The required design is decouple ingestion and processing through SQS with idempotent workers; acceptance depends on evidence that the API acknowledges durable work quickly and delayed workers process queued items without duplicate effects.

B would retry every queued request without idempotent handling, causing duplicate backend work. This is not a defensible answer to the stated constraint (A monolithic queue processor couples request acknowledgement to a slow backend and causes frequent timeouts): asynchronous queuing isolates throughput and availability of upstream and downstream tiers. The required design is decouple ingestion and processing through SQS with idempotent workers; acceptance depends on evidence that the API acknowledges durable work quickly and delayed workers process queued items without duplicate effects.

D would increase synchronous HTTP timeout while the backend remains slow and coupled. This is not a defensible answer to the stated constraint (A monolithic queue processor couples request acknowledgement to a slow backend and causes frequent timeouts): asynchronous queuing isolates throughput and availability of upstream and downstream tiers. The required design is decouple ingestion and processing through SQS with idempotent workers; acceptance depends on evidence that the API acknowledges durable work quickly and delayed workers process queued items without duplicate effects.

A reports an inadequate or failed acceptance condition: A volatile in-memory queue accepts work but loses jobs when the instance restarts. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that the API acknowledges durable work quickly and delayed workers process queued items without duplicate effects. The underlying design requirement is asynchronous queuing isolates throughput and availability of upstream and downstream tiers.

B reports an inadequate or failed acceptance condition: Replayed SQS jobs produce duplicate backend effects instead of one idempotent operation. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that the API acknowledges durable work quickly and delayed workers process queued items without duplicate effects. The underlying design requirement is asynchronous queuing isolates throughput and availability of upstream and downstream tiers.

C reports an inadequate or failed acceptance condition: The API waits on the backend and still times out under its previous synchronous contract. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that the API acknowledges durable work quickly and delayed workers process queued items without duplicate effects. The underlying design requirement is asynchronous queuing isolates throughput and availability of upstream and downstream tiers.

B would create interface endpoints without private DNS or endpoint security groups allowing application HTTPS. This is not a defensible answer to the stated constraint (A migrated service needs private API access to S3 and Secrets Manager without NAT reliance): endpoint types differ by service; NAT is not necessary for supported private endpoint access. The required design is use gateway endpoints for suitable S3 routes and interface endpoints for Secrets Manager with private DNS; acceptance depends on evidence that service calls succeed from isolated private subnets and egress paths remain within policy.

C would use an S3 gateway endpoint alone and assume it also exposes Secrets Manager privately. This is not a defensible answer to the stated constraint (A migrated service needs private API access to S3 and Secrets Manager without NAT reliance): endpoint types differ by service; NAT is not necessary for supported private endpoint access. The required design is use gateway endpoints for suitable S3 routes and interface endpoints for Secrets Manager with private DNS; acceptance depends on evidence that service calls succeed from isolated private subnets and egress paths remain within policy.

D would configure public NAT egress for Secrets Manager despite the no-NAT requirement. This is not a defensible answer to the stated constraint (A migrated service needs private API access to S3 and Secrets Manager without NAT reliance): endpoint types differ by service; NAT is not necessary for supported private endpoint access. The required design is use gateway endpoints for suitable S3 routes and interface endpoints for Secrets Manager with private DNS; acceptance depends on evidence that service calls succeed from isolated private subnets and egress paths remain within policy.

A reports an inadequate or failed acceptance condition: The endpoint's private DNS is disabled and application DNS resolves a public AWS API address. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that service calls succeed from isolated private subnets and egress paths remain within policy. The underlying design requirement is endpoint types differ by service; NAT is not necessary for supported private endpoint access.

C reports an inadequate or failed acceptance condition: S3 calls succeed privately but the Secrets Manager API remains unreachable without NAT. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that service calls succeed from isolated private subnets and egress paths remain within policy. The underlying design requirement is endpoint types differ by service; NAT is not necessary for supported private endpoint access.

D reports an inadequate or failed acceptance condition: Secrets Manager calls still use a NAT gateway despite the endpoint-only requirement. This result does not establish the specified outcome for the current architecture. The decisive test must demonstrate that service calls succeed from isolated private subnets and egress paths remain within policy. The underlying design requirement is endpoint types differ by service; NAT is not necessary for supported private endpoint access.