docs(dgdr): clarify rapid fallback gates (#11040)

Signed-off-by: Dr. Stefan Schimanski <sschimanski@nvidia.com>

Author: sttts

docs/components/profiler/profiler-guide.md

@@ -78,14 +78,18 @@ flowchart TD
 
 ### Rapid
 
-Uses AIC's performance simulation to estimate optimal configurations without deploying real engines. Completes in ~30 seconds.
+Uses AIC's performance simulation to estimate optimal configurations without
+deploying real engines. Completes in ~30 seconds; see the
+[AIC support matrix](https://ai-dynamo.github.io/aiconfigurator/support-matrix/).
 
 ```yaml
 searchStrategy: rapid
 ```
 
 - Supports all backends: vLLM, SGLang, TensorRT-LLM
-- If the model/hardware/backend combination is not supported by AIC, falls back to a naive config (memory-fit TP calculation)
+- Falls back to a naive config (memory-fit TP calculation) only after DGDR
+  accepts the GPU SKU. Fallback sizing depends on AIC system metadata and does
+  not add support for additional GPU SKUs.
 - No GPU resources consumed during profiling
 
 ### Thorough

docs/kubernetes/model-deployment-guide.md

@@ -159,9 +159,10 @@ GPU resources consumed during profiling.
 - Your GPU SKU is in the [AIC support matrix](#aic-support-matrix)
 
 **Limitations:**
-- If AIC does not support your model/hardware/backend combination, the profiler
-  falls back to a naive memory-fit config (basic TP calculation) which may not
-  be optimal.
+- Fallback to a naive memory-fit config only applies after DGDR accepts
+  `hardware.gpuSku`. Fallback sizing depends on AIC system metadata and does
+  not add support for additional GPU SKUs; unsupported model/GPU/backend
+  combinations can fall back but may be suboptimal.
 - Simulated results may differ from real-hardware performance for unusual
   configurations.
 
@@ -177,23 +178,28 @@ benchmarks with AIPerf. Takes 2–4 hours.
 
 **Use thorough when:**
 - Tuning for production and you need the most optimal configuration
-- Your hardware is not supported by AIC (e.g., PCIe GPUs)
 - You want measured rather than simulated performance data
 
 **Constraints:**
 - **Disaggregated mode only** — thorough does not run aggregated configurations.
 - **`backend: auto` is not supported** — you must specify `vllm`, `sglang`, or
   `trtllm`. The DGDR will be rejected if you use `auto` with `thorough`.
+- **Still requires AIC generator support** — thorough measures candidates on
+  real GPUs, but uses AIC to enumerate candidates and generate the final DGD.
 - **Requires GPU resources** — the profiler deploys real inference engines on
   your cluster during profiling.
 
 ## DGDR Detail: AIC Support Matrix
 
-The rapid strategy relies on AIC performance models. AIC currently supports:
+The rapid strategy relies on AIC system metadata and performance models. Check
+the [AIC support matrix](https://ai-dynamo.github.io/aiconfigurator/support-matrix/)
+for the latest support. Measured profiling still needs AIC system and generator
+support to enumerate and render candidates; rapid also needs performance support
+for the exact model/GPU/backend combination.
 
 ### GPU SKUs
 
-| Supported (rapid) | Not Yet Supported (use thorough) |
+| Supported (rapid) | Not supported by rapid |
 |---|---|
 | H100 SXM | V100 (SXM/PCIe) |
 | H100 PCIe | T4 |
@@ -208,8 +214,8 @@ The rapid strategy relies on AIC performance models. AIC currently supports:
 
 > [!NOTE]
 > Some rapid-mode SKUs use AIC estimate-only data until measured profiles are
-> available. Use `searchStrategy: thorough` when you need hardware-measured
-> profiling for an estimate-only or unsupported SKU.
+> available. Use `searchStrategy: thorough` for measured profiling only when
+> DGDR accepts the SKU and AIC has system and generator support for it.
 
 When specifying GPU SKUs manually, use lowercase underscore format (e.g.,
 `h100_sxm`, not `H100-SXM5-80GB`). See the