diff --git a/src/saveload/saveload.cpp b/src/saveload/saveload.cpp
index d67aa8c10d..0046c698d3 100644
--- a/src/saveload/saveload.cpp
+++ b/src/saveload/saveload.cpp
@@ -2736,25 +2736,26 @@ struct SaveLoadFormat {
 	byte min_compression;                 ///< the minimum compression level of this format
 	byte default_compression;             ///< the default compression level of this format
 	byte max_compression;                 ///< the maximum compression level of this format
+	bool no_threaded_load;                ///< unsuitable for threaded loading
 };
 
 /** The different saveload formats known/understood by OpenTTD. */
 static const SaveLoadFormat _saveload_formats[] = {
 #if defined(WITH_LZO)
 	/* Roughly 75% larger than zlib level 6 at only ~7% of the CPU usage. */
-	{"lzo",    TO_BE32X('OTTD'), CreateLoadFilter<LZOLoadFilter>,    CreateSaveFilter<LZOSaveFilter>,    0, 0, 0},
+	{"lzo",    TO_BE32X('OTTD'), CreateLoadFilter<LZOLoadFilter>,    CreateSaveFilter<LZOSaveFilter>,    0, 0, 0, true},
 #else
-	{"lzo",    TO_BE32X('OTTD'), nullptr,                            nullptr,                            0, 0, 0},
+	{"lzo",    TO_BE32X('OTTD'), nullptr,                            nullptr,                            0, 0, 0, false},
 #endif
 	/* Roughly 5 times larger at only 1% of the CPU usage over zlib level 6. */
-	{"none",   TO_BE32X('OTTN'), CreateLoadFilter<NoCompLoadFilter>, CreateSaveFilter<NoCompSaveFilter>, 0, 0, 0},
+	{"none",   TO_BE32X('OTTN'), CreateLoadFilter<NoCompLoadFilter>, CreateSaveFilter<NoCompSaveFilter>, 0, 0, 0, false},
 #if defined(WITH_ZLIB)
 	/* After level 6 the speed reduction is significant (1.5x to 2.5x slower per level), but the reduction in filesize is
 	 * fairly insignificant (~1% for each step). Lower levels become ~5-10% bigger by each level than level 6 while level
 	 * 1 is "only" 3 times as fast. Level 0 results in uncompressed savegames at about 8 times the cost of "none". */
-	{"zlib",   TO_BE32X('OTTZ'), CreateLoadFilter<ZlibLoadFilter>,   CreateSaveFilter<ZlibSaveFilter>,   0, 6, 9},
+	{"zlib",   TO_BE32X('OTTZ'), CreateLoadFilter<ZlibLoadFilter>,   CreateSaveFilter<ZlibSaveFilter>,   0, 6, 9, false},
 #else
-	{"zlib",   TO_BE32X('OTTZ'), nullptr,                            nullptr,                            0, 0, 0},
+	{"zlib",   TO_BE32X('OTTZ'), nullptr,                            nullptr,                            0, 0, 0, false},
 #endif
 #if defined(WITH_LIBLZMA)
 	/* Level 2 compression is speed wise as fast as zlib level 6 compression (old default), but results in ~10% smaller saves.
@@ -2762,9 +2763,9 @@ static const SaveLoadFormat _saveload_formats[] = {
 	 * The next significant reduction in file size is at level 4, but that is already 4 times slower. Level 3 is primarily 50%
 	 * slower while not improving the filesize, while level 0 and 1 are faster, but don't reduce savegame size much.
 	 * It's OTTX and not e.g. OTTL because liblzma is part of xz-utils and .tar.xz is preferred over .tar.lzma. */
-	{"lzma",   TO_BE32X('OTTX'), CreateLoadFilter<LZMALoadFilter>,   CreateSaveFilter<LZMASaveFilter>,   0, 2, 9},
+	{"lzma",   TO_BE32X('OTTX'), CreateLoadFilter<LZMALoadFilter>,   CreateSaveFilter<LZMASaveFilter>,   0, 2, 9, false},
 #else
-	{"lzma",   TO_BE32X('OTTX'), nullptr,                            nullptr,                            0, 0, 0},
+	{"lzma",   TO_BE32X('OTTX'), nullptr,                            nullptr,                            0, 0, 0, false},
 #endif
 };
 
@@ -3176,7 +3177,9 @@ static SaveOrLoadResult DoLoad(LoadFilter *reader, bool load_check)
 	}
 
 	_sl.lf = fmt->init_load(_sl.lf);
-	_sl.lf = new ThreadedLoadFilter(_sl.lf);
+	if (!fmt->no_threaded_load) {
+		_sl.lf = new ThreadedLoadFilter(_sl.lf);
+	}
 	_sl.reader = new ReadBuffer(_sl.lf);
 	_next_offs = 0;