G84 dwell at end of retract

Hi, I am new to tapping with a CNC, so maybe this is just me being ignorant, but I have noticed on a dry run of my first tapping attempt that the machine does not pause to let the spindle reverse and get up to speed after retracting from a tap.  The issue is that it rapids over to the next hole, and starts to feed down into the next hole all before the spindle is up to speed in the forward direction. 

Should there not be a pause of the specified dwell time at the top of the hole as well as the bottom to let the machine catch up?  My only workaround is to force the machine to have a high feed plane so it takes longer to go from hole to hole. 

I am using a 5hp vector duty baldor to drive the spindle on my bridgeport Boss, with a hitachi VFD and braking resistors. Its not the fastest, but it seems like it has a reasonable spindle acceleration rate.  I have the spindle at speed setup so that it anticipates the rate of acceleration of the spindle correctly, but it looks to me as if this is not utilized in spindle reversing.  I was surprised that it is not.

I realize these issues are probably requiring updates to the linuxcnc program, but would improve the functionality, and might be worth the effort.  OR maybe everyone these days is just using rigid tap except for us bridgeport guys.

I do have spindle soft start setup so the machine waits to start cutting for it to be up to speed. It works, and I have scoped it to check functionality.

# ---spindle soft start---
loadrt limit2 names=spindle-ramp
loadrt near names=spindle-at-speed
addf spindle-ramp servo-thread
addf spindle-at-speed servo-thread
setp spindle-ramp.maxv 2100
net spindle-cmd <= spindle.0.speed-out-abs => spindle-ramp.in
net spindle-ramped <= spindle-ramp.out
net spindle-cmd => spindle-at-speed.in1
net spindle-ramped => spindle-at-speed.in2
net spindle-ready <= spindle-at-speed.out => spindle.0.at-speed

Not an LCNC guru, but does it make a difference in how you program multiple holes?  i.e. a separate G84 word/line with a G80 at the end of each hole, then a new M3 Sxxx & G84?

Just curious, because if it works as expected when each hole is programmed as a completely separate operation then I'd think LCNC is ignoring the second (and other) spindle-at-speed check while the canned cycle is active.