bug#34821: discard_input_tty does not discard pending input, resulting in garbage inserted into the buffer

[Platon Pronko]

I guess that something in (terminal-init-xterm) hooks results in some buffer 
switch event being fired, that triggers the second return from read_char(). But 
since (terminal-init-xterm) and read_key_sequence() run concurrently, we have a 
race condition - if the terminal responds quickly enough, second buffer switch 
does not trigger quick enough and this problem happens.

I verified this by adding (sleep-for 0.1) to (defun xterm--query) or (defun 
terminal-init-xterm) - this resulted in 100% reproduction of the bug (thanks 
for the directions, by the way - my hands already hurt after trying to trigger 
that 1-in-10 bug so many times).

Maybe trying to rely on terminal being slow with the response is not the best 
solution? Even if I find a cause of the buffer switch event, we won't have a 
way to ensure that it always arrives before the response to SDA query (because 
multithreading).