The reason for this is, because the last change in the file structure is buggy, it was only written the message length of the buffer to the file and missed to add the length for the timestamp.
For fixing this problem it would be possible to insert something like

len += sizeof(quint64);

after:

int len = mavlink_msg_to_send_buffer(buf+sizeof(quint64), &message);

But I think a consistent file structure has more adavatages for a logfile over a more compressed structure.
E.g. it would be necessary to go through all packages to proceed a jump during a replay.
Additional the LogPlayer was not updated during the last 7 month since the logfile structure was changed (even it was buggy)...

If filesize is really a problem I would recommend to change the logfile handling in a way of writting a temporary consistend
logfile and compress it afterwards (QCompress) like done by many programs

Refactored part of MAVLinkProtocol::receiveBytes() as I was trying to debug within it. Nothing major, though I did remove an unnecessary loop.

Code changes to reflect new header information (sysid, compid, timestamp) in protobuf messages. Added protobuf message timeout; old messages will not be visualized. Fixed getParameterValue in ParamManager; the method can segfault if the component id or parameter id is not valid.

Enabled visualization of point cloud data via Protocol Buffers serialization and extended_mavlink_message.

Fixed roll angle in Google earth, still not smooth. Fixed a number of threading issues. Fixed all custom widget related bugs. Improved MAVLink decoder. Reduced CPU usage substantially across all views