It is a more or less general FDM problem, especially once printing speeds above 40mm/s are involved.
For the (older) Bambu printers the flimsy hotend is a severe limitation.
Not just in terms of how restrictive it is but also in terms of its ability to heat the filament sufficiently IN TIME.
Just skip it unless you really want to know lots of boring details...
In the old days printers were slow and hotends came with rather beefy heaters - I had a printer for 3mm filament with TWO 75W heating cartridges in the block…
Today though all is about speed and saving weight…
I have no problem printing PLA at the max 500mm/s my P1S allows and still get really decent results with simple prints.
But I also use a high flow hotend…
And I learned some most vital lessons about the quality of the Bambu P1 hotends…
But lets talk about the reasons behind your issue…
To print at high speeds the extruder has to be able to supply the filament at a matching temp AND viscosity…
Things like ramping up and down for directional changes these days is mainly focussed on the motion hardware capabilities but NOT the actual printing.
For this someone came up with Pressure Advance settings.
Bambu calls it the K-Factor…
It is a way to more or less accurately calculate how fast the extruder has to spin when making directional changes and by how much the angle of this change affects things.
A circle is easy, a star with 18 points not so much…
The traditional Bambu issue here is this nowhere in the Wiki you will find something dominant telling about the important relationships of temp, speed, flow ratio, k-factor, filament properties and what not…
While Bambu got better with the profiles for their supported filaments and the default print profiles it does not mean they got it perfect, nor that there are no difference between batches…
Is is a good (more or less) AVERAGE - good for general use…
You only get the best POSSIBLE print if your print settings and filament calibration are spot on - or as close as possible, given the limitations.
Before I fry your brain cells through total boredom, try some basic math>
The filament has a diameter of 1.75mm, the nozzle 0.4mm.
Calculate the resulting area and use this to calculate the extruded VOLUME for some layer heights from 0.2mm to 0.08mm…
You might be surprised by the huge volume changes 
Ok, why does this happen and then so badly?
An ideal wall is a CLOSED WALL.
But once you have holes or other openings the nozzle has to make a sharp turn and go back - same story on the other side of the opening.
If you can print at 200mm/s using those basic test prints it won’t always mean you can make a reverse turn within a 3mm wall at those speeds.
The hardware won’t allow it and with that the extruder HAS TO slow down with the movement.
If you had a relatively long straight your nozzle got to a decent speed, then it has to slow down while the hotend overshoots thanks to the pressure inside.
The acceleration value define the ‘ramping’ - going lower means the heads starts to slow down early and more gentle and in return the pressure change in the melting chamber is not as severe.
But one should only try these and some other ‘tricks’ AFTER a proper filament calibration and confirming said calibration with some quick but meaningful test prints.
If in doubt then just don’t trust your nickel-plated eyesight and use a magnifying glass to see when those patches change from having a gap to just showing closed top layer lines - select the patch AFTER the one that still has a gap here and there.
For the second round you select the patch with the best surface and overlap quality AFTER the one that still has some underfill between the lines.
If 0 already has gaps or the last patch still has no gaps then start over.
The k-factor can be tricky and you SHOULD use a good magnifying glass and preferably a marker pen.
With the pattern check for when the corner changes from sharp to bulged.
Look for where the corner looks the sharpest on the inside and outside while not having a gap between the lines.
If you want to print with fine layers do the entire thing again but specify a range between those two good looking corners and use a higher resolution to get more corners to check.
If you do overhang and bridging calibrations that require you change the nozzle temp please double check your calibration values still match as a temp change affects them.
Still not good enough ?
Cut out just the layers around the problem zone and print it at 50% speed.
If that provide a great improvement you something to go by and if it looks great you know your calibration works.
Now for the fun part - finding how fast you can go…
Since holes and other wall openings are the main issue it makes sense to just create a test wall with a few holes in it.
Like in a similar thickness as in your problem print but making sure you have at least one long straight before and after the last hole.
This way you get two end corners and with three holes 6 additional ‘corners’ to judge.
The outer ones, if looking too bad usually get much better if the speed discontinuity area is reduced until they get nice and sharp.
If you have to go below 90 for this to happen consider reducing you default print speeds first.
Those holes, round or square, add a lower print speed as the head can reach full speed between them - giving different results at times.
If you have to adjust acceleration values to still allow for high print speeds where possible then start low and work your way up.
Increment in ‘jumps’ and when those problem areas start to look bad go back in smaller steps until they are acceptable.
Just to complete the confusion> Why do some filaments struggle more than others in this department?
For starter because they come with different calibration profiles - obvious really
But PLA, PEGT, Nylon ABS, not to mention filled filaments of all sorts also mean they come with different physical properties.
Requiring different nozzle temps or scrawling speeds is a dead give away but most people don’t really think about how this translates into FDM.
Take PLA…
It has a certain temp range from becoming sticky enough to bond to basically turning into a liquid - a rather slim range.
On the other hand PETG or Nylon can turn quite soft and become sort of tacky long before this tackiness is enough for a proper layer bond.
It is not just the temp here but actual chemistry and things like surface energy…
Boring stuff…
With all this is seems obvious that those filament also behave quite different in the melting chamber of the hotend…
PLA in an otherwise high speed print might struggle badly with elevated nozzle temps required for those speeds during the last layers or if there is island areas on the layers with little surface area.
And Nylon or PETG might turn into a nightmare if you usually printed lots of small things by the dozen to then help a friend with something big that allows for this print head to actually reach the set speeds…
So frustrating when a 500g+ Nylon print ends in the bin because the layers won’t hold…
And all that just because the filament overcooked or did not have enough time in the melting chamber to reach the required viscosity…
Which brings me to the final and most frustrating part of all this…
Speed, as in how fast AND EVENLY your extruder spins is the main factor for all those wall ‘defects’ not even the best calibration make go disappear.
Think about this for a little moment>
If you want to print a delicate figurine or head scan you print it SLOW to get the best possible detail.
If you want to print a simple cylindrical vase you do this in vase mode but apart from that you do it FAST.
It should be logical to think that if a model suddenly changes the level of detail that this would have some affect on the print.
Modern slicers, printers and of course hyped advertising made us think we only have to press a button and the rest just happens.
The old generation printers were considered fast if they were able to print PLA at 60 to 70mm/s…
And for them everyone accepted that at those ‘insane’ speed the quality might suffer a bit.
Sadly the evolution in the slicing engines was overshadowed by all printer manufacturers going 32bit for their processors.
More speed, more tinkering, more fun…
All while those faster processors also meant we could use better stepper drivers to get a higher accuracy and more control.
Now speed is already a relative thing for the advertising, features, multi-material support, laser support, cutting support, drawing support, whatever support,…
As if we really need a woolen, egg laying and milk providing pig…
No surprise then that many users, especially new ones forget about physics and real world limitations.
IF all is calibrated well and the at least the temp and print speed are within acceptable level prints will come out as good as the hardware and slicer allows for.
That does not mean however that these theoretically best results are always possible.
Layer times and sudden directional changes are still the holy grail to figure out so to speak.
Ideally we would print with an even extrusion rate as only this provides consistent results throughout.
But there isn’t really much out there in the slicing world for printing with a consistent or at least controlled flow rate.
Same, by the way, for non-planar slicing.
We call it 3D printing despite all we ever do is to print in layered 2D ROFL
Anyway, long story short>
Calibrate your filament and temps then check at 50% speed if problematic print areas turn out much better.
If so then first try to improve thing through the speed discontinuity area settings and if not enough by finding a balance between max speed and fastest acceleration to get the required detail level.
Do yourself a huge favour and start from a standard print profile.
Only change one setting at a time and save the profile with a meaningful name so you have something to go back to in case you end up confused or thing won’t work after a few changes.
Aim to create at least 3 profiles>
General use, high speed, high detail/complex model.
Trust me when I say that especially for larger models it really helps being able able to judge what the model requires and to accept the limitations.
It is far better to wait another hour for a print to complete than to take it out and find it in bad quality of with bad layer adhesion.
If you have supplies of consistent quality it is well worth the time and material to create the best possible print profile and best possible filament calibrations.
Even if a batch arrives with a different quality you can adjust what is required and save it as a new profile without having to start from scratch.
