Filed from the IronCache pre-implementation coverage audit (2026-06-13): no existing issue adequately owned this.
Why this is needed
Compatible is the top tenet and the only proof a client is actually compatible is running that driver's own integration suite (and real apps) against IronCache across a pinned client-version matrix over RESP2 and post-HELLO-3 RESP3. This catches connection-pooling, pipelining, reconnect/RESET, CLIENT INFO parsing, RESP3-push handling, and error-string pattern-matching breakages a byte-diff-against-Valkey harness never exercises because no real client is in the loop. #95's acceptance contains only the one-line aspiration 'unmodified clients run green' with no harness, version matrix, or design; #17 greps driver SOURCE for error-text matching (static analysis), not running them; #97 diffs RESP bytes against pinned servers with no real driver. No issue owns running the suites or the version matrix the lens calls out. Directly defends the top tenet.
Context
Relates to / partially overlaps #95. Part of the vision EPIC #1.
Why this is needed
Compatible is the top tenet and the only proof a client is actually compatible is running that driver's own integration suite (and real apps) against IronCache across a pinned client-version matrix over RESP2 and post-HELLO-3 RESP3. This catches connection-pooling, pipelining, reconnect/RESET, CLIENT INFO parsing, RESP3-push handling, and error-string pattern-matching breakages a byte-diff-against-Valkey harness never exercises because no real client is in the loop. #95's acceptance contains only the one-line aspiration 'unmodified clients run green' with no harness, version matrix, or design; #17 greps driver SOURCE for error-text matching (static analysis), not running them; #97 diffs RESP bytes against pinned servers with no real driver. No issue owns running the suites or the version matrix the lens calls out. Directly defends the top tenet.
Context
Relates to / partially overlaps #95. Part of the vision EPIC #1.