After running BUFGMUX example on Nexys Video with limited grid I obtained the bitstream, which is not working correctly on the hardware. The diff_fasm step returns no differences between the FASMs generated by the vivado and symbiflow. In addition, a bitstream generated by the vivado in the diff_fasm step works correctly.

Here is the output of thebit2fasm.py script used on vivado diff_fasm bitstream:

{ unknown_bit = "00401b9a_43_4", unknown_segment = "0x00401b80", unknown_segbit = "26_1380" }
{ unknown_bit = "00401b9a_55_20", unknown_segment = "0x00401b80", unknown_segbit = "26_1780" }
{ unknown_bit = "00401b9a_98_14", unknown_segment = "0x00401b80", unknown_segbit = "26_3150" }

This leads to HCLK_R_X137Y130 in the xc7a200tffg1156-1 tilegrid.json.
The features related to the HCLK_R are generated by the 058-pip-hclk fuzzer.