As seen today on IRC (#gnumeric):

16:01 < brick> here is a strange side effect, =countif(I$37:I$106,"<=E28") 
               won't evaluate the E28... hrm
16:03 <@TreMobyl> indeedy
16:03 <@TreMobyl> bugzilla it
16:03 < brick> well, i think i can see why the string doesn't get variable 
               evaluation... perhaps there is a concatenaty way to assemble the 
               string in advance
16:05 < brick> gnumeric can do string manipulation, i assume... reading up...
16:06 <@TreMobyl> yes
16:06 <@TreMobyl> is still one too many hoops for my taste
16:06 <@TreMobyl> I mean, the formula you pasted should work if it's in another 
                  setting.
16:06 < brick> i'm reluctant to call it a bug, actually
16:07 <@TreMobyl> true, it is a string.
16:07 < brick> the "<=val" API is sketchy and causes this side effect
16:07 <@TreMobyl> Although the IF would imply that it's not a string but a 
                  conditional expression, which it (arguably) ought to be.
16:07 < brick> definitely easier in some programming languages
16:05 < brick> gnumeric can do string manipulation, i assume... reading up...
16:06 <@TreMobyl> yes
16:06 <@TreMobyl> is still one too many hoops for my taste
16:06 <@TreMobyl> I mean, the formula you pasted should work if it's in another 
                  setting.
16:06 < brick> i'm reluctant to call it a bug, actually
16:07 <@TreMobyl> true, it is a string.
16:07 < brick> the "<=val" API is sketchy and causes this side effect
16:07 <@TreMobyl> Although the IF would imply that it's not a string but a 
                  conditional expression, which it (arguably) ought to be.
16:07 < brick> definitely easier in some programming languages
16:07 < brick> totally agreed
16:07 <@TreMobyl> I suspect that this is Excel-Compatible, though.
16:07 < brick> that could be a nice enhancement
16:08 <@TreMobyl>     * This function is Excel compatible.
16:08 <@TreMobyl>                 
16:08 < brick> :/
16:08 <@TreMobyl> looks like we would need a G_COUNTIF
16:09 <@TreMobyl> confirmed with XL2003
16:10 < brick> ahh
16:10 <@TreMobyl> behaves exactly like gnumeric does.  No string as an error, 
                  no cell expansion in the string.
16:10 <@TreMobyl> and MS Office won't let me quit until I fix the error.  LAME!
16:10 < brick> i got the whole question answered with 
               =countif(I$37:I$106,concatenate("<=",E29))  -  the number of 
               years prior to the fund's meeting a certain level.
16:10 < brick> yikes
16:10 <@TreMobyl> yes, but that is stupid.  :)
16:11 < brick> hah
16:11 <@TreMobyl> should have a g_countif that actuall uses a conditional, not 
                  a string

I therefore propose a new, gnumeric-specific function, namely g_countif where the condition is a *conditional* not a string!