mcl 2.0.1 glitches

[Dimitri Simos <dim@lissys.demon.co.uk>]

I've been converting from 2.0 to 2.0.1 and have a few minor
questions which I hope you could help me with. None of
these is high priority, but they do baffle me.

---1. I create my application with the default #'toplevel-function
since I still want the listener to be there. However, I would prefer
to suppress the "Welcome to Macintosh Common Lisp Version 2.0.1!"
message which comes up. Is there a simple way of doing this?
I do not really understand how to create the appropriate toplevel
loop, but I found by experiment that in the previous MCL version (2.0)
this message went away if I specified:
      :toplevel-function #'toplevel-loop
when doing the save-application.
(It's amazing what your users will try, eh?)
However, trying the above in 2.0.1 crashes MCL and the system, so for
the moment I do not give the keyword argument.

---2. This is a problem I've been noticing for some time:
(format t "~,2f" 0.006)  gives 0.01 as expected, but:
(format t "~,2f" 0.004)  gives 0.004  ! Similarly,
(format t "~,2f" 0.000001) gives 0.000  etc.
I thought that there could never be more than the specified
number of digits after the decimal. Or am I reading Steele
CLtL2 pp.588-589 wrong?

---3. Loading the "ccl:examples;boyer-moore" file seems to give
a slower performance than it did under version 2.0.  When I specify
a double wildcard (e.g. "HD:a folder:*:*") the system takes a long
time being quiet before I hear any heavy disk activity. Is this
normal?

---4. I have some problems with Balloon help. I want my application to
run without a Help file, so I moved it elsewhere. My application then kept 
popping up dialogs looking for it, so I changed the argument from t to nil in:
(defmethod help-resource-file ((a application))
  (get-doc-string-file t))
purely by guesswork and it seemed to work. Is this ok? However, trying
to get balloon help on the "Search.." dialog then gave nasty messages, so
I changed  (error "Malformed help-spec: ~s" restype) to simply (cancel).
Also the function  help-tehandle  sometimes complained of a non-string
argument.  Furthermore, the problem of the missing descenders seems
to be still there sometimes, although it may be my fault for messing with
a file that I do not understand.   Sorry for this flurry of ignorance,
this is rather low priority as I said.

Finally, I've been testing my application on a PowerMac and despite
the disappointing speed I have to say it's proven very robust and
problem-free (just the known clash with Virtual Memory). 
Well done indeed to Apple for such a faithful emulation.
Let's see what awaits the really wonderful MCL now..

Thanks in advance,
Dimitri Simos