How might I work with a data.frame where each physical row represents several logical rows?

 ...
Perhaps something like this:

# function to read the values in 'values':
 parse_values <- function(x) {scan(text= gsub( "\\[|\\]","",x), sep=",") }

# the apply function reads line-by-line
 new_dat <- apply(test_data, 1, function(d) data.frame( as.list(d[!names(d)  %in% "values"]), nvals <- parse_values(d['values']) ) )

...
# Could suppress the report from scan by adding quiet = TRUE
# now take this list of 4 line data.frames and "rbind" them
# If you wanted these to remain character you would use stringsAsFactors=FALSE in the data.frame call

new_df <- do.call("rbind", new_dat)

...

(I will also end up collecting all of the records for a given timestamp
and hostname, and creating one very wide record with all of the data
from the set of records thus found.  I already have (yes, Perl) code to
do this -- though if there's a reasonable way to avoid that, I'm
interested.)

I thought you wanted the data in long form.

...