RTU issues related to RTS timing

[chstrunk]

When trying to use RTU on Raspberry Pi (using 64-Bit Raspberry Pi OS, os-issue 2025-10-01, based on Debian Trixie) I ran into trouble like bad CRC and timeout. Then when watching the RTS signal with an oscilloscope I saw that it was held active around twice as long as it should be, so after the RasPi has sent out a request the RTU is kept active too long, and thus the start of the device's answer is missed.
After looking into the sources I was able to fix that in modbus-rtu.c inside function _modbus_rtu_send be replacing
usleep(ctx_rtu->onebyte_time * req_length + ctx_rtu->rts_delay);
with
tcdrain(ctx->s);
usleep(ctx_rtu->rts_delay);
My guess is that the preceding write(ctx->s, req, req_length) already internally waits until all (or almost all) bytes are transmitted and thus waiting another ctx_rtu->onebyte_time * req_length makes the RTS signal much too long.
Replacing this with tcdrain should IMHO have the same effect if bytes were still buffered on a system, but should return immediately if everything was transmitted already.
Do you agree to this analysis and do you think this is a good solution?