Evaluate the impact on MongoDB of https://github.com/wiredtiger/wiredtiger/pull/14958, which addresses WT-9074 by bounding how long a snapshot waits for a stalled transaction ID allocation, retrying allocations a snapshot has marked, and removing the allocation wait from the oldest-transaction scan.

Test plan

• Compare identical MongoDB builds and configurations using the PR base revision versus the patched revision.
• Run read-heavy, write-heavy and mixed workloads, including multi-document transactions, across increasing concurrency and CPU oversubscription.
• Measure throughput, p50/p95/p99 latency, CPU utilization and transaction abort/retry rates. Record the new txn_id_alloc_retry statistic where exposed.
• Exercise transaction correctness, snapshot isolation, prepared transactions, replication, failover and recovery using the relevant MongoDB test suites.
• Repeat benchmarks and record hardware, revisions, configurations and run-to-run variation.

Acceptance criteria

• Baseline-versus-patch results and supporting artifacts are published.
• Determine whether tail latency improves under scheduling pressure and whether ordinary workloads regress.
• Investigate correctness failures or performance differences beyond baseline variability, and recommend whether further changes or testing are needed.