[script] Fix parallel mode scripts with delays cancelling each other (#10324)

esphome/core/base_automation.h

@@ -5,6 +5,8 @@
 #include "esphome/core/hal.h"
 #include "esphome/core/defines.h"
 #include "esphome/core/preferences.h"
+#include "esphome/core/scheduler.h"
+#include "esphome/core/application.h"
 
 #include <vector>
 
@@ -158,7 +160,16 @@ template<typename... Ts> class DelayAction : public Action<Ts...>, public Compon
   void play_complex(Ts... x) override {
     auto f = std::bind(&DelayAction<Ts...>::play_next_, this, x...);
     this->num_running_++;
-    this->set_timeout("delay", this->delay_.value(x...), f);
+
+    // If num_running_ > 1, we have multiple instances running in parallel
+    // In single/restart/queued modes, only one instance runs at a time
+    // Parallel mode uses skip_cancel=true to allow multiple delays to coexist
+    // WARNING: This can accumulate delays if scripts are triggered faster than they complete!
+    // Users should set max_runs on parallel scripts to limit concurrent executions.
+    // Issue #10264: This is a workaround for parallel script delays interfering with each other.
+    App.scheduler.set_timer_common_(this, Scheduler::SchedulerItem::TIMEOUT,
+                                    /* is_static_string= */ true, "delay", this->delay_.value(x...), std::move(f),
+                                    /* is_retry= */ false, /* skip_cancel= */ this->num_running_ > 1);
   }
   float get_setup_priority() const override { return setup_priority::HARDWARE; }
 

esphome/core/scheduler.cpp

@@ -65,14 +65,17 @@ static void validate_static_string(const char *name) {
 
 // Common implementation for both timeout and interval
 void HOT Scheduler::set_timer_common_(Component *component, SchedulerItem::Type type, bool is_static_string,
-                                      const void *name_ptr, uint32_t delay, std::function<void()> func, bool is_retry) {
+                                      const void *name_ptr, uint32_t delay, std::function<void()> func, bool is_retry,
+                                      bool skip_cancel) {
   // Get the name as const char*
   const char *name_cstr = this->get_name_cstr_(is_static_string, name_ptr);
 
   if (delay == SCHEDULER_DONT_RUN) {
     // Still need to cancel existing timer if name is not empty
+    if (!skip_cancel) {
       LockGuard guard{this->lock_};
       this->cancel_item_locked_(component, name_cstr, type);
+    }
     return;
   }
 
@@ -97,7 +100,9 @@ void HOT Scheduler::set_timer_common_(Component *component, SchedulerItem::Type
   if (delay == 0 && type == SchedulerItem::TIMEOUT) {
     // Put in defer queue for guaranteed FIFO execution
     LockGuard guard{this->lock_};
+    if (!skip_cancel) {
       this->cancel_item_locked_(component, name_cstr, type);
+    }
     this->defer_queue_.push_back(std::move(item));
     return;
   }
@@ -150,9 +155,11 @@ void HOT Scheduler::set_timer_common_(Component *component, SchedulerItem::Type
     return;
   }
 
-  // If name is provided, do atomic cancel-and-add
+  // If name is provided, do atomic cancel-and-add (unless skip_cancel is true)
   // Cancel existing items
+  if (!skip_cancel) {
     this->cancel_item_locked_(component, name_cstr, type);
+  }
   // Add new item directly to to_add_
   // since we have the lock held
   this->to_add_.push_back(std::move(item));

esphome/core/scheduler.h

@@ -21,8 +21,13 @@ struct RetryArgs;
 void retry_handler(const std::shared_ptr<RetryArgs> &args);
 
 class Scheduler {
-  // Allow retry_handler to access protected members
+  // Allow retry_handler to access protected members for internal retry mechanism
   friend void ::esphome::retry_handler(const std::shared_ptr<RetryArgs> &args);
+  // Allow DelayAction to call set_timer_common_ with skip_cancel=true for parallel script delays.
+  // This is needed to fix issue #10264 where parallel scripts with delays interfere with each other.
+  // We use friend instead of a public API because skip_cancel is dangerous - it can cause delays
+  // to accumulate and overload the scheduler if misused.
+  template<typename... Ts> friend class DelayAction;
 
  public:
   // Public API - accepts std::string for backward compatibility
@@ -184,7 +189,7 @@ class Scheduler {
 
   // Common implementation for both timeout and interval
   void set_timer_common_(Component *component, SchedulerItem::Type type, bool is_static_string, const void *name_ptr,
-                         uint32_t delay, std::function<void()> func, bool is_retry = false);
+                         uint32_t delay, std::function<void()> func, bool is_retry = false, bool skip_cancel = false);
 
   // Common implementation for retry
   void set_retry_common_(Component *component, bool is_static_string, const void *name_ptr, uint32_t initial_wait_time,

tests/integration/fixtures/parallel_script_delays.yaml

@@ -0,0 +1,45 @@
+esphome:
+  name: test-parallel-delays
+
+host:
+
+logger:
+  level: DEBUG
+
+api:
+  actions:
+    - action: test_parallel_delays
+      then:
+        # Start three parallel script instances with small delays between starts
+        - globals.set:
+            id: instance_counter
+            value: '1'
+        - script.execute: parallel_delay_script
+        - delay: 10ms
+        - globals.set:
+            id: instance_counter
+            value: '2'
+        - script.execute: parallel_delay_script
+        - delay: 10ms
+        - globals.set:
+            id: instance_counter
+            value: '3'
+        - script.execute: parallel_delay_script
+
+globals:
+  - id: instance_counter
+    type: int
+    initial_value: '0'
+
+script:
+  - id: parallel_delay_script
+    mode: parallel
+    then:
+      - lambda: !lambda |-
+          int instance = id(instance_counter);
+          ESP_LOGI("TEST", "Parallel script instance %d started", instance);
+      - delay: 1s
+      - lambda: !lambda |-
+          static int completed_counter = 0;
+          completed_counter++;
+          ESP_LOGI("TEST", "Parallel script instance %d completed after delay", completed_counter);

tests/integration/test_automations.py

@@ -89,3 +89,73 @@ async def test_delay_action_cancellation(
         assert 0.4 < time_from_second_start < 0.6, (
             f"Delay completed {time_from_second_start:.3f}s after second start, expected ~0.5s"
         )
+
+
+@pytest.mark.asyncio
+async def test_parallel_script_delays(
+    yaml_config: str,
+    run_compiled: RunCompiledFunction,
+    api_client_connected: APIClientConnectedFactory,
+) -> None:
+    """Test that parallel scripts with delays don't interfere with each other."""
+    loop = asyncio.get_running_loop()
+
+    # Track script executions
+    script_starts: list[float] = []
+    script_ends: list[float] = []
+
+    # Patterns to match
+    start_pattern = re.compile(r"Parallel script instance \d+ started")
+    end_pattern = re.compile(r"Parallel script instance \d+ completed after delay")
+
+    # Future to track when all scripts have completed
+    all_scripts_completed = loop.create_future()
+
+    def check_output(line: str) -> None:
+        """Check log output for parallel script messages."""
+        current_time = loop.time()
+
+        if start_pattern.search(line):
+            script_starts.append(current_time)
+
+        if end_pattern.search(line):
+            script_ends.append(current_time)
+            # Check if we have all 3 completions
+            if len(script_ends) == 3 and not all_scripts_completed.done():
+                all_scripts_completed.set_result(True)
+
+    async with (
+        run_compiled(yaml_config, line_callback=check_output),
+        api_client_connected() as client,
+    ):
+        # Get services
+        entities, services = await client.list_entities_services()
+
+        # Find our test service
+        test_service = next(
+            (s for s in services if s.name == "test_parallel_delays"), None
+        )
+        assert test_service is not None, "test_parallel_delays service not found"
+
+        # Execute the test - this will start 3 parallel scripts with 1 second delays
+        client.execute_service(test_service, {})
+
+        # Wait for all scripts to complete (should take ~1 second, not 3)
+        await asyncio.wait_for(all_scripts_completed, timeout=2.0)
+
+        # Verify we had 3 starts and 3 ends
+        assert len(script_starts) == 3, (
+            f"Expected 3 script starts, got {len(script_starts)}"
+        )
+        assert len(script_ends) == 3, f"Expected 3 script ends, got {len(script_ends)}"
+
+        # Verify they ran in parallel - all should complete within ~1.5 seconds
+        first_start = min(script_starts)
+        last_end = max(script_ends)
+        total_time = last_end - first_start
+
+        # If running in parallel, total time should be close to 1 second
+        # If they were interfering (running sequentially), it would take 3+ seconds
+        assert total_time < 1.5, (
+            f"Parallel scripts took {total_time:.2f}s total, should be ~1s if running in parallel"
+        )