Read the sequences into a DNAStringSet

library(Biostrings)
url <- "http://www.rosaceae.org/sites/www.rosaceae.org/files/Prunus_persica.main_genome.scaffolds.fasta"
peachgenome <- read.DNAStringSet(url)

then do a 'vectorized' count.

vcountPattern("TATAAAA", peachgenome)

This takes about 6 seconds for your 8 strings, or for all of them(!)

> system.time(x <- vcountPattern("TATAAAA", peachgenome[1:8], max.mismatch=1))
   user  system elapsed 
  6.139   0.004   6.148 
> x
[1] 250652 136713 115696 153960  95002 153177 117959 113009
> system.time(x <- vcountPattern("TATAAAA", peachgenome, max.mismatch=1))
   user  system elapsed 
  6.376   0.008   6.391

You do not use sapply like you did above. The construct is much simpler:

sapply(peachgen,function(x) {countPattern(TATA,x,max.mismatch=1)})

This line:

myobs <- paste("peachgenx$scaffold_", 1:8, sep = "")

creates a vector of strings:

"peachgenx$scaffold_1" "peachgenx$scaffold_2" ... "peachgenx$scaffold_8"

You are then using sapply/countPattern to match TATA to those strings, not to the list of sequences. Unsurprisingly, the count = 0, since the string "peachgenx$scaffold_1" does not contain "TATAAAAA".

You need to run lapply or sapply on the list of sequences, stored in peachgenome:

lapply(peachgenome, function (x) { countPattern(TT, x[2]$seq, max.mismatch = 1) })
# OR
sapply(peachgenome, function (x) { countPattern(TT, x[2]$seq, max.mismatch = 1) })