diff --git a/adaptive_scheduler/_executor.py b/adaptive_scheduler/_executor.py
index df098461..19a0f91f 100644
--- a/adaptive_scheduler/_executor.py
+++ b/adaptive_scheduler/_executor.py
@@ -10,7 +10,7 @@
 from concurrent.futures import Executor, Future
 from dataclasses import asdict, dataclass, field
 from pathlib import Path
-from typing import TYPE_CHECKING, Any, Literal, NamedTuple
+from typing import TYPE_CHECKING, Any, NamedTuple
 
 from adaptive import SequenceLearner
 
@@ -82,7 +82,7 @@ class TaskID(NamedTuple):
 class SlurmTask(Future):
     """A `Future` that loads the result from a `SequenceLearner`."""
 
-    __slots__ = ("executor", "task_id", "_state", "_last_size", "min_load_interval")
+    __slots__ = ("executor", "task_id", "_state", "_last_size", "min_load_interval", "_load_time")
 
     def __init__(
         self,
@@ -94,8 +94,8 @@ def __init__(
         self.executor = executor
         self.task_id = task_id
         self.min_load_interval: float = min_load_interval
-        self._state: Literal["PENDING", "RUNNING", "FINISHED", "CANCELLED"] = "PENDING"
         self._last_size: float = 0
+        self._load_time: float = 0
 
     def _get(self) -> Any | None:  # noqa: PLR0911
         """Updates the state of the task and returns the result if the task is finished."""
@@ -126,6 +126,8 @@ def _get(self) -> Any | None:  # noqa: PLR0911
         self._last_size = size
 
         learner.load(fname)
+        self._load_time = time.monotonic() - now
+        self.min_load_interval = max(1.0, 20 * self._load_time)
         self.executor._run_manager._last_load_time[idx_learner] = now
 
         if idx_data in learner.data: