while i dont run X on the machines i see this effect on (since 2.0.x>3, i
believe--dont hold me to that), i often see problems with screen. trying to
create a new window is effectively a nop. and, up until lately, i
solved it by somehow opening another telnet window to localhost in order
to allocate the pty in a normal fashion. however, i recently discovered
that given enough persistence, you can overcome the bug merely by dishing
out the create requests fast and furious...between the 5th and 20th try,
whatever was broke is fixed. this is unfortunately as unpredictable a
result as is the bug itself---no apparently commonality between the
environments under whcih it happens (screen running as a user or root on
either of two machines, no common tty involved, no nothing...)
wish i could reliably reproduce it to do the binary kernel search. but, it
definitely happens on 2.0.10 and 2.0.18-21.