bug in http / autoproxy?

I have a server that stops working now and then. It gets stuck somewhere and it does not respond to any requests anymore. This is a public facing server and it receives a lot of bad requests and attacks.
With many puts statements sprinkled here and there, I have traced it down to this error message:
TCL READ VARNAME
can't read "state(sock)": no such element in array
     while executing
"fileevent $state(sock) writable {}
I don't have this line anywhere in my code. But they are present in http and autoproxy (with version info: http 2.9.1, autoproxy 1.7).
In http, it occurs in http::reset and http::Connect.  The one in http::reset is in a catch statement so it should be OK, but the one in http::Connect is not, and I suspect this is where the error comes from.
In autoproxy, it occurs in ::autoproxy::tunnel_write, again without a catch.
I am not really familiar with either of these packages. Putting a catch around these calls seems simple enough but I wonder if an expert can provide a more robust fix.