CLAUDE: pose stage non-exclusive with the render product (pose = part of the RT cycle)

Andrey's ruling: the pose stage normally ALWAYS runs (it is part of the RT
cycle, not a mutually-exclusive test) - the last run had rend_test ON which
silently skipped the pose stage (old precedence) and left TILE-CALIB unbuilt.
Now: pose_test runs FIRST, rend_test render follows when also ON, cond_test
stays standalone (only when nothing else ran). Dialog/tooltips updated
('CUAS RT render (full product)').

Verified meanwhile (POSE-11 renders): frozen per-sequence DC cut the
RT-RENDER horizon wobble 234 -> 42.4 counts pk-pk (5.5x), exactly to the
oracle DBG floor - the residual at 1-rev period is the shared FPN ghost;
FPN.tiff still Jun-1 vintage because cuas_calc_fpn was OFF (the '(Re)calculate
FPN' checkbox) - the rot_period=164.9 recompute has not actually run yet.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

src/main/java/com/elphel/imagej/cuas/rt/CuasRT.java

@@ -45,9 +45,11 @@ import ij.ImagePlus;
  * Top-level flow (keep {@link #run} READABLE - it is the map of the RT mode):
  *   1. INITIALIZATION - transform the borrowed data (center ingest, per-scene
  *      instances) and fix missing calibration (per-sensor photometric).
- *   2. DIAGNOSTICS  - optional exclusive test modes (rend_test / pose_test /
- *      cond_test). They REPLACE the production run and return; they must never
- *      leak into the production flow nor grow the top method.
+ *   2. DIAGNOSTICS  - optional test/prototype stages (pose_test / rend_test /
+ *      cond_test). The POSE stage is part of the RT cycle and normally always
+ *      runs (Andrey 07/06/2026); the render product follows it when enabled;
+ *      cond_test is standalone. They replace the production detection and
+ *      return; they must never leak into the production flow nor grow the top.
  *   3. PRODUCTION   - the per-input(scene) processing loop and target detection.
  *      MVP target shape: for each input scene, a multi-threaded read of the 16
  *      sensor /jp4/*.tiff (or their hardware-stream equivalents later) ->
@@ -330,45 +332,49 @@ public class CuasRT {
 	}
 
 	/**
-	 * Optional EXCLUSIVE diagnostic modes - each replaces the production run. Precedence
-	 * (unchanged from the OpticalFlow inline era): rend_test > pose_test > cond_test.
-	 * Keep new tests HERE - the production path and the top method must not grow.
-	 * @return true if a diagnostic ran (caller returns without production processing).
+	 * Optional diagnostic/prototype stages. NOT mutually exclusive anymore (Andrey
+	 * 07/06/2026): the POSE stage is part of the RT cycle and normally ALWAYS runs
+	 * (checkbox retained until the production loop consumes the poses directly); the
+	 * render product runs AFTER it when also enabled. cond_test remains a standalone
+	 * diagnostic (runs only when nothing else did). Keep new tests HERE - the
+	 * production path and the top method must not grow.
+	 * @return true if any stage ran (caller returns without production detection).
 	 */
 	private static boolean diagnostics(
 			final CLTParameters clt_parameters,
 			final QuadCLT       master_CLT,
 			final QuadCLT []    quadCLTs,
 			final int           debugLevel) {
-		if (clt_parameters.curt.rend_test) {
-			// MINIMAL RT ingest+render certification: per scene raw /jp4/ ->
-			// conditionSceneToGpu -> virtual-grid render at BORROWED stored pose + ERS
-			// rates -> -CUAS-RT-RENDER hyperstack, comparable slice-by-slice to the
-			// -CUAS-INDIVIDUAL/MERGED-CUAS-DBG oracle products. Certify/debug THIS chain
-			// first, only then the pose test (Andrey 07/05/2026). By Claude on 07/05/2026.
-			System.out.println("===== CUAS RT render test (curt_rend_test): ingest+render sequence vs oracle DBG =====");
-			CuasRender.testRenderSequence(
+		boolean ran = false;
+		if (clt_parameters.curt.pose_test) {
+			// The POSE stage (part of the RT cycle): re-generate per-scene 3-angle poses
+			// against the virtual-center reference, ascending, prediction-seeded, single
+			// pass on the final combo DSI. Output: -POSE-RT-TEST.csv + fitted-vs-stored
+			// summary (truth = scenes_poses restored from INTERFRAME corr-xml).
+			// By Claude on 07/03/2026, from Andrey's design.
+			System.out.println("===== CUAS RT pose stage (curt_pose_test): per-scene 3-angle fit vs virtual center =====");
+			CuasPoseRT.testPoseSequence(
 					clt_parameters, // CLTParameters clt_parameters,
 					master_CLT,     // QuadCLT       center_CLT,
 					quadCLTs,       // QuadCLT []    quadCLTs,
 					debugLevel);    // int           debugLevel
-			return true;
+			ran = true;
 		}
-		if (clt_parameters.curt.pose_test) {
-			// RT pose-adjustment prototype: re-generate per-scene 3-angle poses against the
-			// virtual-center reference, ascending, prediction-seeded, single pass on the
-			// final combo DSI. Output: -POSE-RT-TEST.csv + fitted-vs-stored summary (truth =
-			// scenes_poses restored from INTERFRAME corr-xml). By Claude on 07/03/2026,
-			// from Andrey's design.
-			System.out.println("===== CUAS RT pose test (curt_pose_test): per-scene 3-angle fit vs virtual center =====");
-			CuasPoseRT.testPoseSequence(
+		if (clt_parameters.curt.rend_test) {
+			// Full-render product (after the pose stage when both are ON): per scene raw
+			// /jp4/ -> conditionSceneToGpu -> virtual-grid render at BORROWED stored pose +
+			// ERS rates -> -CUAS-RT-RENDER hyperstack, comparable slice-by-slice to the
+			// -CUAS-INDIVIDUAL/MERGED-CUAS-DBG oracle products. By Claude on 07/05/2026,
+			// non-exclusive per Andrey 07/06/2026.
+			System.out.println("===== CUAS RT render (curt_rend_test): ingest+render sequence vs oracle DBG =====");
+			CuasRender.testRenderSequence(
 					clt_parameters, // CLTParameters clt_parameters,
 					master_CLT,     // QuadCLT       center_CLT,
 					quadCLTs,       // QuadCLT []    quadCLTs,
 					debugLevel);    // int           debugLevel
-			return true;
+			ran = true;
 		}
-		if (clt_parameters.curt.cond_test) {
+		if (!ran && clt_parameters.curt.cond_test) {
 			// Conditioning/calibration diagnostic: raw /jp4/ baseline (no photometric/FPN/
 			// conditioning), saved as -CUAS-PERSENSOR-RAW for side-by-side compare with the
 			// conditioned -CUAS-PERSENSOR (saved by the calibration step when curt.calib is
@@ -392,9 +398,9 @@ public class CuasRT {
 					master_CLT.getGPUQuad(), // GpuQuad with the current 16-sensor TD
 					master_CLT,              // debug-stack save target (-CUAS-TDAVG-CHECK)
 					debugLevel);
-			return true;
+			ran = true;
 		}
-		return false;
+		return ran;
 	}
 
 	/**

src/main/java/com/elphel/imagej/tileprocessor/CuasRtParameters.java

@@ -28,7 +28,7 @@ public class CuasRtParameters {
 	public boolean  pose_corr_save =  false; // DEBUG: save the per-scene 2D correlations vs the virtual center in the pixel domain (-POSE-RT-CORR2D: z=scenes, tile grid of 16x16 cells, last LMA cycle) - lean engine only. // By Claude on 07/04/2026
 	public boolean  pose_img_save =   false; // DEBUG: save the composite scenes actually correlated - imclt render of the consolidated 16-sensor TD average at the converged pose (-POSE-RT-COMPOSITE: z=scenes; NaN outside task tiles) + the virtual-center reference render (-POSE-RT-CENTER-REF). Lean engine only. // By Claude on 07/04/2026
 	public boolean  pose_stored =     false; // DEBUG: decouple rendering/measurement from the LMA - use the STORED (oracle vintage) pose for every scene, no adjustment; one leanMeasure per scene feeds -POSE-RT-HYPER/-CORR2D/-COMPOSITE. The composite must then match the oracle CUAS-MERGED-CUAS-DBG renders tile-for-tile. Lean engine only. // By Claude on 07/04/2026
-	public boolean  rend_test =       false; // MINIMAL RT ingest+render test (CuasRender.testRenderSequence, takes precedence over pose test): per scene raw /jp4/ -> conditionSceneToGpu -> virtual-grid render at BORROWED stored pose+ERS rates -> -CUAS-RT-RENDER hyperstack [s00..s15, merged][scenes], comparable to -CUAS-INDIVIDUAL/MERGED-CUAS-DBG. Certify THIS before the pose test. // By Claude on 07/05/2026
+	public boolean  rend_test =       false; // RT full-render product (CuasRender.testRenderSequence): per scene raw /jp4/ -> conditionSceneToGpu -> virtual-grid render at BORROWED stored pose+ERS rates -> -CUAS-RT-RENDER hyperstack [s00..s15, merged][scenes], comparable to -CUAS-INDIVIDUAL/MERGED-CUAS-DBG. NON-exclusive: runs AFTER the pose stage when both are ON (the pose stage is part of the RT cycle - Andrey 07/06/2026). // By Claude on 07/05/2026
 	public boolean  dbg_fpixels =     true;  // save the ORACLE debug renders during ingest (prepareFpixels): -CUAS-INDIVIDUAL-CUAS-DBG (an EXTRA per-sensor render pass, ~11GB) + -CUAS-MERGED-CUAS-DBG (save only - the merged render is the production input regardless). Was hardcoded. Turn OFF to save the render time and disk once the RT chain (-CUAS-RT-RENDER) is the trusted full render. // By Claude on 07/05/2026
 	public double   fz_inter =        10000.0; // RT fat zero, INTER (pose/motion: scene x virtual-center correlation). DECOUPLED from the legacy gpu_fatz*/AUX switches (which stay untouched for non-RT/FOPEN). Value is for a single physical-pair amplitude; FZ ~ amplitude^2 in the GPU normalize - scale down x4..x16 for consolidated-average applications as needed. 10000 = the proven oracle INTER operating point (the lean code erroneously used the INTRA value before). // By Claude on 07/05/2026
 	public double   fz_intra =        2000.0; // RT fat zero, INTRA (ranging/disparity: sensor-pair correlations within one scene). Reserved for the RT ranging path; same decoupling/scaling notes as fz_inter. // By Claude on 07/05/2026
@@ -114,8 +114,8 @@ public class CuasRtParameters {
 				"DEBUG: save the rendered composite (TD-averaged, grid-transformed) scenes correlated against the virtual center (-POSE-RT-COMPOSITE, z=scenes) + the center reference render (-POSE-RT-CENTER-REF). Lean engine only.");
 		gd.addCheckbox ("Pose test stored poses (no LMA)",           this.pose_stored, // By Claude on 07/04/2026
 				"DEBUG: measure/render every scene at its STORED pose, skip the LMA entirely - decouples the task/render/correlation chain from the solver. Lean engine only.");
-		gd.addCheckbox ("CUAS RT render test (ingest+render)",       this.rend_test, // By Claude on 07/05/2026
-				"MINIMAL standalone test (precedence over pose test): raw jp4 -> conditioning -> virtual-grid render at borrowed stored poses/ERS; saves -CUAS-RT-RENDER [s00..s15+merged][scenes] to compare against the -CUAS-*-DBG oracle products.");
+		gd.addCheckbox ("CUAS RT render (full product)",             this.rend_test, // By Claude on 07/05/2026
+				"Full RT-chain render: raw jp4 -> conditioning -> virtual-grid render at borrowed stored poses/ERS; saves -CUAS-RT-RENDER [s00..s15+merged][scenes], comparable to the -CUAS-*-DBG oracle products. NON-exclusive: runs AFTER the pose stage when both are ON.");
 		gd.addCheckbox ("Save oracle DBG renders (11GB)",            this.dbg_fpixels, // By Claude on 07/05/2026
 				"Save -CUAS-INDIVIDUAL-CUAS-DBG (an EXTRA per-sensor render pass, ~11GB) and -CUAS-MERGED-CUAS-DBG during ingest. Turn OFF to save render time and disk once -CUAS-RT-RENDER (RT render test) is the trusted full render.");
 		gd.addNumericField("RT fat zero INTER (pose/motion)",        this.fz_inter, 1,9,"", // By Claude on 07/05/2026