diff --git a/src/dexorder/final_ohlc.py b/src/dexorder/final_ohlc.py
index 7255214..439a3a0 100644
--- a/src/dexorder/final_ohlc.py
+++ b/src/dexorder/final_ohlc.py
@@ -233,6 +233,7 @@ class OHLCFileSeries:
         self.symbol = symbol
         self.series_start: Optional[datetime] = None  # timestamp of the first datum in the series
         self.write_offset: Optional[int] = None
+        self.start_time_bytes: Optional[bytes] = None  # if populated, these bytes are written at the start of the line on flush()
         self.last_flush: Optional[tuple[datetime,dec]] = None
         self.quote_file = None
         self.dirty_files = set()
@@ -245,31 +246,7 @@ class OHLCFileSeries:
 
 
     def update(self, time: datetime, price: dec):
-        #
-        # load quote file
-        #
-        if self.quote_file is None:
-            quote_filename = self.quote_filename
-            try:
-                self.quote_file = open(quote_filename, 'br+')
-                # load existing quote file
-                line = self.quote_file.read().decode('ascii')
-            except FileNotFoundError:
-                os.makedirs(os.path.dirname(quote_filename), exist_ok=True)
-                self.quote_file = open(quote_filename, 'bw')
-                line = None
-            if line:
-                start, old_time, old_price = line.split(',')
-                self.series_start = from_timestamp(int(start))
-                # position the write cursor at the start of the second column so we can write the latest quote quickly
-                self.write_offset = len(start) + 1  # after the start time bytes and comma
-                self.quote_file.seek(self.write_offset)
-                self.last_flush = from_timestamp(int(old_time)), dec(old_price)
-            else:
-                # initialize new quote file with our own series_start time
-                self.quote_file.write((str(timestamp(time)) + ',').encode('ascii'))
-                self.write_offset = 0
-                self.last_flush = None
+        self._load(time)
 
         #
         # forward-fill OHLC files that would otherwise be empty/skipped
@@ -295,6 +272,39 @@ class OHLCFileSeries:
             file.update(time, price)
             self.dirty_files.add(file)
 
+
+    def _load(self, time):
+        #
+        # load quote file
+        #
+        if self.quote_file is None:
+            quote_filename = self.quote_filename
+            try:
+                self.quote_file = open(quote_filename, 'br+')
+                # load existing quote file
+                line = self.quote_file.read().decode('ascii')
+            except FileNotFoundError:
+                os.makedirs(os.path.dirname(quote_filename), exist_ok=True)
+                self.quote_file = open(quote_filename, 'bw')
+                line = None
+            try:
+                start, old_time, old_price = line.split(',')
+            except ValueError:
+                line = None
+            if line:
+                # noinspection PyUnboundLocalVariable
+                self.series_start = from_timestamp(int(start))
+                # position the write cursor at the start of the second column so we can write the latest quote quickly
+                self.write_offset = len(start) + 1  # after the start time bytes and comma
+                self.quote_file.seek(self.write_offset)
+                # noinspection PyUnboundLocalVariable
+                self.last_flush = from_timestamp(int(old_time)), dec(old_price)
+            else:
+                # initialize new quote file with our own series_start time
+                self.start_time_bytes = f'{timestamp(time)},'.encode('ascii')
+                self.write_offset = 0
+                self.last_flush = None
+
     def flush(self):
         if self.quote is None:
             log.warning('OHLCFileSeries was flushed without having any updated data.')
@@ -314,12 +324,16 @@ class OHLCFileSeries:
         #
         try:
             start_of_column = self.quote_file.tell()
+            if self.start_time_bytes:
+                self.quote_file.write(self.start_time_bytes)
+                self.start_time_bytes = None
             self.quote_file.write(f'{ts},{price:f}\n'.encode('ascii'))
             self.quote_file.truncate()
             self.quote_file.flush()
             self.quote_file.seek(start_of_column)
         except IOError:
             log.exception(f'While saving quote file {self.quote_filename}')
+            raise
 
         # remember the last flush point so we can forward-fill next time
         self.last_flush = self.quote