Is there a problem with Digital Unix 4.0G, ASE 12.0, Ct-Lib 11.5 ?

We are experiencing problems where, rarely, a CT-Lib connection will be
stuck in
a recv call (waiting for an answer) forever. Most of the time (99.9+ %)
of the time,
the connections work fine. We develop applications on both HP and
Compaq and are
only seeing this on Digital Unix 4.0G using ASE 12.0 (EBF/SWR 9502)

This problem of seemingly "forgotten" connections did not happen on the
exact same code when
using ASE 11.5 and the same O/S. The connections are all CT-Lib
from the same host as the server using standard CT-Lib 11.5 calls.
Also, the exact
same processes running on the HP boxes do not get the problem.

The condition does not seem to be system/database load dependent. We
run both
our HP and Compaq systems with 2GB of memory and two database engines.

The queries in question are almost always very simple queries that work
fine 99.9% of the time.

One other note is that we are seeing some bursts of errors in the log
file of the sort:

kernel dstartio: I/O request repeatedly delayed; block number: 362713,
vdn: 101

Our async i/o parameters on the server and the O/S are :

Group: O/S Resources

Parameter Name Default Memory Used Config Value Run Value

max async i/os per engine 2147483647 0 1024
max async i/os per server 2147483647 0 2048
o/s file descriptors 0 0 0 4096
tcp no delay 0 0 1 1

The DBA's feel that the i/o per server and engine should be like this
we run two engines.

In any case, the dstart i/o error's do not always correlate with the
not getting an answer from the server.

Sybase Support claims there are no known issues with the combination of
server/lib/platform, but I'm convinced something is subtly wrong here.

Any help/similar experience would be greatly appreciated.

select @@version info:
Adaptive Server Enterprise/ 9502 ESD 2/Digital UNIX/OSF1
V4.0D/1668/64bit/OPT/Thu Feb 22 06:16:43 2001
Ct-Lib 11.5
Digital Unix 4.0G