The 16650 and the 16750 apparently do hardware RTS/CTS.  Plug compatible with
the 16550.  I'll let you know if I find a single source and get a chance to
play with them.

Heres some comments from someone running one of those expensive commercial
unixes:

>[ Technofeeb ] Message 93 (35 left): Thu Feb 1, 1996  12:49pm
>From: feeb me (spcecdt@deepthought.armory.com)
>Subject: 16[67]50
>21 lines (?) 
>
>
>They sound very useful to me.  The new OS release I'm running reports receiver
>overruns.  They started showing up immediately.  Not constant, but enough that
>I thought I should do something about it.  It's probably not occuring for
>interactive use, just during file transfers, but there are too many fragile
>implementations of transfer protocols out there to let it slide.  I held off on
>doing anything about it until I had upgraded the hw (486/66->P100).  That
>helped (by reducing interrupt latency), but I was still getting overruns.  Next
>I tried reducing all of the 16550 FIFO trigger levels from 8 to 4.  That
>reduced it further.  Still overruns.  Didn't bother reducing triggers from 4 to
>2; that would increase the buffer-after-interrupt from 12 to 14 which isn't
>likely to make a difference in the long-latency cases that are causing this.
>So, I finally switched all 9 of my v32bis modem lines from 38400 to 19200. 
>That still leaves some room for modem compression, but it's annoying that I had
>to do that.  I left the 3 ZyXEL and v34 lines at 38400 (the ZyXELs have a
>proprietary 19200 modulation that some users take advantage of, and I wanted to
>leave room for compression on top of that).  I rarely see overruns any more
>(last time was almost a week ago), but I'd really like to have all the lines
>at 38400 (or higher), AND have the trigger level set high to reduce interrupt
>load in streaming-input cases...