At least one who noticed it. I have put a comment on Massdrop about firmware bugs of the XD-05, and someone recommended to post my findings here as well. The unit behaves differently between USB and digital input SPDIF:
Fed via USB: PCM 1 and 3 are Short Delay Sharp, PCM 2 and 4 are Short Delay Slow. Changing from PCM 1 to 2 'Slow' is activated for less than half a second, then the unit activates SD Slow. The same happens when switching from PCM 4 to 1, 'Sharp' gets active for less than half a second.
Fed via SPDIF: SD sharp is active all the time, no matter what is selected. Here also other filters can be seen for a very short time after switching through PCM 1 to 4.
Whatever else you thought you heard was just imagination. My unit is new, and the simple proof that all units behave like this ever since can be found on a russian website which hosts lots of XD-05 measurements. They show the exact same (wrong) impulse responses for the filter types with PCM 1 to 4.
http://reference-audio-analyzer.pro/en/report/dac/xduoo-xd-05-pcm1-spdif.php
http://reference-audio-analyzer.pro/en/report/dac/xduoo-xd-05-pcm2.php
http://reference-audio-analyzer.pro/en/report/dac/xduoo-xd-05-pcm3.php
http://reference-audio-analyzer.pro/en/report/dac/xduoo-xd-05-pcm4.php
But there is more. I have to add the SRC feature as buggy as it also shows a flaw. Playing back 44 kHz with SRC off, it seems the SRC is not off. The signal at the analog output includes high-frequency distortion which is not normal and should not be there. This is immediately solved when using the SRC, so upsampling to 192 kHz as standard operational mode for listening to SPDIF is recommended.
Now I don't want to send back this unit as it is a real bargain and working quite well in most other aspects. But I do want those sloppy chinese developers to show responsibility and fix these problems by a firmware update!
44 kHz playback via SPDIF
44 kHz playback via USB, or using SPDIF with upsampling