My expeirence is that more compelx ICs are
noticably less reliable than
 simple ones. 
 That's much too simple analysis even if it were correct. Failure rates in
 ICs are primarily down to processing, packaging and handling (and to a
 lesser extent die area), not complexity.  
 
I hate to tell you this, but my comments are based on actual, practical,
observations. Not 'what the theory says'.
In other words, I'ev; had many thousands of things across the bench by
now. And I can assure you that complex ICs, modern ones, old ones,
whatever, are mcuh more likely to be dead than simpler ones.
 
 I would think the 555 had a much better lifetime than any
 microcontrolelr
 therefore. 
 You would be wrong however. Today's ICs have a much better lifetime.
 Statistically today's commercial grade ICs have a significantly better
 lifetime than military grade ICs of the 555 era. 
 
Not based on the ICs I have to replace I'm not...
 The origianl 555 (as opposed to the 7555, etc) is large-ish junciton
 bipolar and thus less likely to suffer static damage than a MOS
 microcotnroller. It also has a much wider supply votaeg range and is
 less
 likely to be damaged by supply problems. 
 That's not a definition of reliability - it's one of robustness. Within
 their specifications the 555 would be less reliable (IMHO of course). 
 
Hmm... I am tlaking abvout the reliabiltiy of the whole system. In other
words, how often do I ahve to fix it? What has failed? It may be that the
IC has failed because it was essentially misused (supply glitches,
whatever), but a device more susceptable to those is likely to lead to
prodcut that fails more often.
 And of coruse if a microcontroller fails, unless you have the program
 to
 put into it, you're stuck. If a 555 fails, well, there's a tube of 50
 in
 my spares box. And I cna go to any number of suppliers anf buy one.
  
 Agreed, but off-topic ;o) 
 
It's actualyl a very major consideration for a lot of the designs I do.
-tony