Is there an efficient way to backward search a large vector in R?

I have a vector of 10+ million elements. I need to find all elements satisfying a given condition A (e.g. X < 2 at rows i %in% c(6,10)). From each of these elements I need to skim the vector backwards and flag all preceding elements while they satisfy condition B (e.g. X < 4 for i %in% c(8:10) and c(5:6)).

For example, given the following X column, I would like the final result to be the flag2 column. I am not interested in elements where B is true if they are not immediately preceding an element satisfying A, therefore row i == 2 has flag2 == 0.

  i  |  X  | flag1 | flag2
---------------------------
  1  |  4  |   0   |   0
  2  |  3  |   0   |   0
  3  |  6  |   0   |   0
  4  |  9  |   0   |   0
  5  |  3  |   0   |   1
  6  |  1  |   1   |   1
  7  |  9  |   0   |   0
  8  |  3  |   0   |   1
  9  |  2  |   0   |   1
 10  |  1  |   1   |   1

The first operation to produce flag1 is simple and very fast:

# locate all occurrences of X < 2
my_data$flag1 = dplyr::case_when(my_data$X < 2 ~ 1, T ~ 0)

I have implemented the second operation with the following for loop, which gives the desired result but is prohibitively time-consuming given the amount of data.

# flag all elements preceding the ones already flagged while they satisfy `X < 4`
my_data$flag2 = my_data$flag1
for(i in nrow(my_data):2){
  if((my_data[i,]$flag2 == 1) & (my_data[i-1,]$X < 4)){ 
    my_data[i-1,]$flag2 = 1
  }
}

Is there any way I could do this more efficiently?