Please be much more specific.
Are you breaking one drill bit at a time,
or all / many drillbits,
are you trying for better reliability,
are you doing lots of parallel workpieces,
or one piece that is very expensive with n x 80+ holes drilled for some reason ?
Like a very long part maybe ?
What is your *goal*.
What is the main *problem*.
One problem is that 3-10$ drill bits break and one is that multi-1000$ workpieces need expensive rework.
See ?
Drilling SS is dead easy.
When You define the problem better solutions more appropriate can be suggested.
E.g.
If the machine has a variable pot, analog control, for speed,
One option.
1. replace it with a multi-detent switch and add suitable resistors and trimmer pots on each line.
You could get a 40-pos switch, for example, and have very repeatable control of speed on each detent.
The trimmer pots would give a great repeatable tuning ability.
Cost == 30$ for 30-40 lines aka nothing, plus maybe 3 hours of fiddling.
Add plugs !
(Plugs of choice. DB25 would work fine, probably. So would ethernet or maybe better shielded ethernet, both cheap and dense in volume).
So you could always use the original system, or return to it, in a few minutes.
--
Speed, volume, and usually but maybe/probably not here - quality (for machinists), are always the constraints.
Let us know about the volume desired, and the key constraint(s) as You see it/them.
Can you check for individual drill TIR ?
Overall TIR avg. ?
If the TIR is very bad, that might be likely, (or not,) then custom epoxy-mounted zero-tir drillbits would maybe fix 99% of your problems.
If the drillbit wobbles in the mount, or the bit is not concentric, etc. then all sorts of cascading exponential errors tend to happen apart from the drillbit itself.
So a very flexible HSS drill might be more forgiving initially vs a better more rigid drill, and then at random times the accumulated stress leads to bits failing be it hss, cobalt, or the more rigid modern drill options.
My humble opinion.
Test first with one drillbit, all other stations empty.
Use a gage pin of drillbit size.
Measure the gage root and tip TIR, movement linearity as the pin (simulated drill) advances.
Compare to drill D, tool total length, drilling depth.
Then repeat the measurements in 2-3 other stations.
You will have actual real data .. that is comparable to cutting tool data.
Supposition:
If you do have a matrix of worn mounts, grabbing drillbits in worn tooling,
at worst 3-jaw chucks well-used or very worn,
imho - it is likely many or a lot of the drills may point wherever, and the axis of advance is sometimes very much not along the theoretical axis of advance.
And thus replacing a drill in the same chuck will often have the drillbit pointing wherever, just due to wear and chance.
There are endless cheap ways of making pot chucks with excellent repeatable TIR, alignment, etc. depending.
Thus YOU need to let us know what You want to try to do in terms of volume, speed, cost, etc.
It is very different if e.g.
1. you want to try to do one total run of maybe 100 sets x 80 matrix and drill-bit breakage is an annoyance alone.
2. you are trying to do 200--xx runs per week for 20 weeks.
If you have time in hours, or some small cash in widgets, or a need for results aka plenty of cash as long as it works, the better critical paths may be very different.
--
Generally, on PM, the willingness to spend 8 work-hours or 1000$ in cash or a minimax of the two is assumed for an issue.
This is an industrial site on how to do things "really well" in a competitive modern environment, either making widgets or production or "really good" or "really cool".
So if someone wants to save say 50$ on 600$ cost in drillbits, on a one-off job, many qualified may not bother to reply, because to us(them) the right answer is "spend the money" and we usually do so every day.
.. the extra cost in money for tooling/stuff is much more profitable than us doing it "better" via an extra 8 work hours at 150$ implied marginal *net* income loss for us.
We actually like and help new guys.
But they must "show willing".