Because after the slice() you only have 1 repetition of each, the order of the commands mathers.

library(tidyverse)
set.seed(2021)
df <- tibble(tt = LETTERS[c(sample(26)[1:20],sample(26)[1:10],sample(26)[1:10],sample(26)[1:10])])
df %>%
slice(1:26) %>%
add_count(tt) %>%
filter(tt %in% c("K", "V"))
#> # A tibble: 2 x 2
#> tt n
#> <chr> <int>
#> 1 K 1
#> 2 V 1
df %>%
add_count(tt) %>%
#slice(1:26) %>%
mutate(fff = fct_lump_min(tt,min = 2, other_level = "under 2")) %>%
arrange(desc(n))
#> # A tibble: 50 x 3
#> tt n fff
#> <chr> <int> <fct>
#> 1 U 4 U
#> 2 I 4 I
#> 3 K 4 K
#> 4 U 4 U
#> 5 I 4 I
#> 6 K 4 K
#> 7 I 4 I
#> 8 K 4 K
#> 9 U 4 U
#> 10 U 4 U
#> # … with 40 more rows

^{Created on 2021-05-21 by the reprex package (v2.0.0)}