fix(sync): bound slow MTP rollout and retained generations

CHANGELOG.md

@@ -5,9 +5,13 @@
 The isolated `trainlog-program-recalage-v1` candidate adds an explicit
 preview-and-confirm cascade for unfinished Program sessions. It retains source
 and actual dates, stores absolute current dates and ceded-slot state, and
-selects a separate Programs V2 Android projection by capability. This candidate
+selects a separate Programs V2 Android projection by capability. The source
-is not merged into the stable installation or deployed to Android.
+candidate remains isolated from the stable checkout while its private desktop
-Rollout packaging now rejects a native build without the embedded Web frontend.
+and compatible Android build are installed for the coordinated rollout.
+Rollout packaging rejects a native build without the embedded Web frontend.
+The private rollout preserves parsed V1 Program number types during Android V2
+upgrade and raises the bounded local MTP operation budget to 90 seconds and
+retained outgoing generation capacity to 16 after interrupted exchanges.
 
 Development opened after the stable v0.1.6 release. This line is restricted to
 observed bug fixes, regression fixes, directly related small UI corrections,

docs/current_state.md

@@ -59,7 +59,7 @@ callbacks complete before the mutable generation reference is published; the
 reference and Android ACK use the same confirmation. Bounded diagnostic logs
 contain only run/generation identities, artifact names, phases and results.
 The desktop worker performs lightweight local correlation checks continuously
-but opens a full MTP pull no more than once every three seconds. A 30-second
+but opens a full MTP pull no more than once every three seconds. A 90-second
 individual adapter budget is distinct from the overall conversation deadline,
 so an absent correlated generation reports a protocol wait timeout while an
 actually blocked adapter reports `transport_timeout`.

docs/sync_exchange.md

@@ -1112,7 +1112,7 @@ named by that validated reference. It does not recursively mirror retained
 Android generations, desktop publications, or unrelated exchange history.
 Local correlation is checked every 50 ms without USB work; complete MTP pulls
 are throttled to a three-second cadence. Each MTP operation has its own
-30-second ceiling while the conversation retains its configured global
+90-second ceiling while the conversation retains its configured global
 deadline. Exhausting the global deadline while waiting for a correlated
 reference reports `timeout waiting for correlated android-generation-v1.json`;
 only an adapter that exceeds its independent operation budget reports

docs/sync_rollout_readiness.md

@@ -110,7 +110,7 @@ These are unsigned development evidence, not deployment authorization.
 
 Do not delete an unacknowledged generation, the last acknowledged generation,
 its previous valid generation, or causal evidence. An ACK timeout is unresolved,
-not rollback. Eight retained outgoing generations exhaust admission and require
+not rollback. Sixteen retained outgoing generations exhaust admission and require
 operator diagnosis; this tranche adds no tombstone or evidence garbage
 collection. A migrated database is recovered from its verified backup, never by
 assuming an older binary reverses its migration.

tests/test_sync_error_classification.py

@@ -128,21 +128,25 @@ class SyncErrorClassificationTest(unittest.TestCase):
             )
 
     def test_transport_timeout_has_a_stable_code_and_bounded_diagnostic(self):
-        timeout = subprocess.TimeoutExpired(["adapter", "push"], 30)
+        operation_budget = sync_peer_worker.MTP_OPERATION_TIMEOUT_SECONDS
+        timeout = subprocess.TimeoutExpired(
+            ["adapter", "push"], operation_budget,
+        )
         with tempfile.TemporaryDirectory() as directory:
             transport = Path(directory) / "transport"
             transport.mkdir()
             with mock.patch.object(sync_peer_worker.subprocess, "run", side_effect=timeout):
                 with self.assertRaisesRegex(
                     RuntimeError,
-                    r"^transport_timeout: MTP push did not finish within 30 seconds$",
+                    rf"^transport_timeout: MTP push did not finish within "
+                    rf"{operation_budget:g} seconds$",
                 ) as caught:
                     sync_peer_worker.run_adapter(
                         Path("/adapter"),
                         "push",
                         "peer_00000000-0000-4000-8000-000000000001",
                         transport,
-                        sync_peer_worker.time.monotonic() + 60,
+                        sync_peer_worker.time.monotonic() + operation_budget + 10,
                     )
         self.assertEqual(sync_orchestrator.failure_code(caught.exception), "transport_timeout")
 

tests/test_sync_peer_worker.py

@@ -58,7 +58,8 @@ class SyncPeerWorkerTest(unittest.TestCase):
         with mock.patch.object(worker.subprocess, "run", return_value=completed) as run:
             worker.run_adapter(
                 Path("/fixed/adapter"), "pull", "peer_fixed",
-                self.transport, 100.0, clock=lambda: 50.0,
+                self.transport, 50.0 + worker.MTP_OPERATION_TIMEOUT_SECONDS + 10.0,
+                clock=lambda: 50.0,
             )
         self.assertEqual(worker.MTP_OPERATION_TIMEOUT_SECONDS, run.call_args.kwargs["timeout"])
 
@@ -112,7 +113,8 @@ class SyncPeerWorkerTest(unittest.TestCase):
             with self.assertRaisesRegex(RuntimeError, "transport_timeout"):
                 worker.run_adapter(
                     Path("/fixed/adapter"), "pull", "peer_fixed",
-                    self.transport, 100.0, clock=lambda: 50.0,
+                    self.transport, 50.0 + worker.MTP_OPERATION_TIMEOUT_SECONDS + 10.0,
+                    clock=lambda: 50.0,
                 )
 
     def test_delayed_correlated_generation_survives_multiple_polls(self):

tools/sync_generation_exchange.py

@@ -51,7 +51,11 @@ MAX_GENERATION = 256 * 1024 * 1024
 MAX_NAME = 64
 MAX_PATH = 240
 MAX_DIAGNOSTIC = 1024
-MAX_RETAINED_PER_PEER = 8
+# WHY: interrupted MTP publications can leave immutable generations awaiting
+# an ACK alongside the two acknowledged lineage members kept for recovery.
+# CONTRACT: admission remains bounded and never erases unresolved generations.
+# INVARIANT: captured, published and waiting generations still consume a slot.
+MAX_RETAINED_PER_PEER = 16
 ID = re.compile(r"^(?:gen|peer|sy)_[0-9a-f]{8}-[0-9a-f]{4}-4[0-9a-f]{3}-[89ab][0-9a-f]{3}-[0-9a-f]{12}$")
 HEX = re.compile(r"^[0-9a-f]{64}$")
 

tools/sync_peer_worker.py

@@ -37,7 +37,11 @@ CAPS = {
 
 ACK_RECOVERY_CAP = "generation-ack-recovery-v1"
 
-MTP_OPERATION_TIMEOUT_SECONDS = 30.0
+# WHY: a verified personal phone can need more than 30 seconds to publish an
+# immutable Programs generation over MTP. CONTRACT: this bounds each adapter
+# call independently of the configured overall conversation deadline.
+# INVARIANT: an adapter overrun remains an ambiguous transport failure.
+MTP_OPERATION_TIMEOUT_SECONDS = 90.0
 MTP_POLL_INTERVAL_SECONDS = 3.0