Is there any case that framing error in UART communication can cause delay in detection of next byte?

I have a strange situation with my implementation that in all other cases runs smoothly, tested with baud rates up to ~250K. During an interaction with a specific peer that is instructed to reset, a framing error is detected when this peer reappears on the line. This is detected as input 00 with Framing Flag raised. The next byte, around 40ms later, seems to be captured half correctly. Meaning that although expected as 0x52, the input is read as 0xF5. Taking into account that the transmission on uart is taking place LSB first, it gives me the impression that the lsb part (0100) is lost and start bit is detected just before the last bit (0) of this part, so that 1010 (5) is considered as the LSB part of the byte and the following high status of the line is considered as the MSB part (1111 or F) plus stop bit, which ends as 0xF5 in the RDR.

I think this one resembles this one.