diff --git a/ld_eventsource/sse_client.py b/ld_eventsource/sse_client.py
index 1ac774e..a2d0548 100644
--- a/ld_eventsource/sse_client.py
+++ b/ld_eventsource/sse_client.py
@@ -111,7 +111,7 @@ def __init__(
 
         self.__connection_client = connect.create_client(logger)
         self.__connection_result: Optional[ConnectionResult] = None
-        self.__connected_time: float = 0
+        self._retry_reset_baseline: float = 0
         self.__disconnected_time: float = 0
 
         self.__closed = False
@@ -248,10 +248,17 @@ def next_retry_delay(self) -> float:
         return self.__next_retry_delay
 
     def _compute_next_retry_delay(self):
-        if self.__retry_delay_reset_threshold > 0 and self.__connected_time != 0:
-            connection_duration = time.time() - self.__connected_time
+        # If the __retry_reset_baseline is 0, then we haven't successfully connected yet.
+        #
+        # In those situations, we don't want to reset the retry delay strategy;
+        # it should continue to double until the retry maximum, and then hold
+        # steady (- jitter).
+        if self.__retry_delay_reset_threshold > 0 and self._retry_reset_baseline != 0:
+            now = time.time()
+            connection_duration = now - self._retry_reset_baseline
             if connection_duration >= self.__retry_delay_reset_threshold:
                 self.__current_retry_delay_strategy = self.__base_retry_delay_strategy
+                self._retry_reset_baseline = now
         self.__next_retry_delay, self.__current_retry_delay_strategy = (
             self.__current_retry_delay_strategy.apply(self.__base_retry_delay)
         )
@@ -287,7 +294,7 @@ def _try_start(self, can_return_fault: bool) -> Optional[Fault]:
                 # If can_return_fault is false, it means the caller explicitly called start(), in
                 # which case there's no way to return a Fault so we just keep retrying transparently.
                 continue
-            self.__connected_time = time.time()
+            self._retry_reset_baseline = time.time()
             self.__current_error_strategy = self.__base_error_strategy
             self.__interrupted = False
             return None
diff --git a/ld_eventsource/testing/test_sse_client_retry.py b/ld_eventsource/testing/test_sse_client_retry.py
index 39783de..fba2fe8 100644
--- a/ld_eventsource/testing/test_sse_client_retry.py
+++ b/ld_eventsource/testing/test_sse_client_retry.py
@@ -1,3 +1,5 @@
+from time import sleep
+
 from ld_eventsource import *
 from ld_eventsource.actions import *
 from ld_eventsource.config import *
@@ -101,6 +103,57 @@ def test_all_iterator_continues_after_retry():
         assert client.next_retry_delay == initial_delay * 2
 
 
+def test_retry_delay_gets_reset_after_threshold():
+    initial_delay = 0.005
+    retry_delay_reset_threshold = 0.1
+    mock = MockConnectStrategy(
+        RespondWithData("data: data1\n\n"),
+        RejectConnection(HTTPStatusError(503)),
+    )
+    with SSEClient(
+        connect=mock,
+        error_strategy=ErrorStrategy.always_continue(),
+        initial_retry_delay=initial_delay,
+        retry_delay_reset_threshold=retry_delay_reset_threshold,
+        retry_delay_strategy=RetryDelayStrategy.default(jitter_multiplier=None),
+    ) as client:
+        assert client._retry_reset_baseline == 0
+        all = client.all
+
+        # Establish a successful connection
+        item1 = next(all)
+        assert isinstance(item1, Start)
+        assert client._retry_reset_baseline != 0
+
+        item2 = next(all)
+        assert isinstance(item2, Event)
+        assert item2.data == 'data1'
+
+        # Stream is dropped and then fails to re-connect, resulting in backoff.
+        item3 = next(all)
+        assert isinstance(item3, Fault)
+        assert client.next_retry_delay == initial_delay
+
+        item4 = next(all)
+        assert isinstance(item4, Fault)
+        assert client.next_retry_delay == initial_delay * 2
+
+        # Sleeping the threshold should reset the retry thresholds
+        sleep(retry_delay_reset_threshold)
+
+        # Which we see it does here
+        item5 = next(all)
+        assert isinstance(item5, Fault)
+        assert client.next_retry_delay == initial_delay
+
+        # And if we don't sleep long enough, it doesn't get reset.
+        sleep(retry_delay_reset_threshold / 2)
+
+        item6 = next(all)
+        assert isinstance(item6, Fault)
+        assert client.next_retry_delay == initial_delay * 2
+
+
 def test_can_interrupt_and_restart_stream():
     initial_delay = 0.005
     mock = MockConnectStrategy(