Query engine cannot distinguish occurrences of the same query string with different requirements #778

Problem

The same PromQL leaf query string can show up more than once in a workload, each time with different requirements. For example:

• It is issued at different intervals (different repetition_delay_ms).
• It runs on its own in one place and as a sub-expression of a larger query in another.
• The occurrences have different accuracy_sla / latency_sla.

The planner can treat these as separate items with separate requirements; since #776, the optimizer key includes t_repeat_ms and both SLAs. That can lead to more than one deployed sketch backing the same query string.

The query engine has no way to tell these occurrences apart when a query arrives. It looks up the QueryConfig by query string alone and uses the first match:

• asap-query-engine/src/engines/simple_engine/mod.rs:576 (find_query_config)
• asap-query-engine/src/engines/simple_engine/promql.rs:238 (find_query_config_promql_structural)

So when several QueryConfig entries share a query string:

• Only the first one is ever used to answer queries. Other sketches are still ingested and stored, but never queried.
• An occurrence with a strict SLA may be answered by a sketch sized for a relaxed SLA, or one built for a different cadence.
• Which entry wins depends on the order of query_configs, which the planner does not guarantee to be stable.

To investigate

• How the workload, planner and engine should identify a query occurrence beyond its string (cadence, SLA, standalone vs. nested).
• Whether the problem appears today in realistic workloads, and how often.