Question 39
Domain 5Why are *two* watermarks (80% soft + 95% hard) preferred over a single threshold?
Correct answer: B
Explanation
Two watermarks create a two-stage control: “80% leaves enough headroom for a graceful exit” so the system can commit, summarize, or capture state before pressure becomes critical. The “95%” hard limit “catches cases where unexpected context spikes blew past the 80% logic,” providing a safety backstop that a single threshold would miss.
Why each option is right or wrong
A. Two thresholds give the model time to consider whether to exit.
Thresholds are for control actions, not giving the model time to deliberate about exiting.
B. 80% leaves enough headroom for a graceful exit (commit, summary, state-of-play). 95% catches cases where unexpected context spikes blew past the 80% logic.
The dual-threshold design is justified by the need for a staged response: at 80% utilization, the system still has roughly 20% capacity to complete a controlled shutdown sequence, including committing state, producing a summary, and preserving context before the buffer becomes tight. The 95% watermark functions as a hard stop under the governing capacity rule, catching abrupt spikes that can occur between checks and would otherwise bypass the earlier trigger; a single threshold would not provide both early warning and a fail-safe backstop.
C. The spec requires exactly two watermarks at 80% and 95%.
Two thresholds are a design choice for resilience, not a universally required fixed specification.
D. Two thresholds let the user pick which one to honour.
Watermarks are system guardrails; users do not choose which safety threshold applies.