you can recreate the problem in the following way: execute an rcp of a single file from a solaris2.6 machine to a linux5.2 client in a loop. the rcp will eventually hang. if I kill inetd on the redhat5.2 machine and then restart it the rcp's resume. I can rcp another file from a different client while the rcp from the first machine is hung up. the file I am transferring is not large (~37kB).
Check your /var/log/messages for errors. Inetd has code that turns off any service that is respawning to fast. If you need a higher rate of connections, then modify your inetd.conf to permit that (I've forgotten details).
thanks for your suggestions. I reran my test and got this error message in /var/log/messages: Aug 3 14:57:09 lunchbox inetd[17857]: shell/tcp server failing (looping), service terminated I will look into how to change the respawn rate in inetd.conf