I'm still just testing everything out before I try writing a whole song, but the problem with the WAV sweep wasn't even MIDI, I was messing with the sweep with the GBC buttons and it wasn't really doing anything

I asked because it seems to make a difference. The less I use the noise channel the better total performance I get.
Is your "music" heavy in one particular channel?  Are you using the noise channel very much?

Oh wait, no it doesn't

Other channels, I'm not just going to test one at a time, I want to use it for music, you know?

I was thinking about that, but the SP uses the same port for link cables and audio

from my experience, seems to work pretty damn well with a gba sp as well if you feel like trying that

Are you triggering just the WAV channel, or and other channels too at the same time?

I dunno I was messing with the CGB and it seemed like the sweep function on the WAV channel didn't work very well

Just for the sake of completeness I thought I would post my results of performance tweaking.

First and foremost Trash is right, if you plan on any live performance, CGB is the best way to go with mGB.

My testing included 4 midi interfaces.  Motu Express 128, RME Fireface, Motu Express XT, EMU 0404 and a cheap midi to usb.
I tested 3 Arduinoboy interfaces.  Trash's original hardware config(UNO), a NeX diy kit (Arduinopromini) and my Synthboy+(also Arduinopromini).
I used CuBase and drew a single note every 2 seconds across 30 bars.  I did this for all four midi channels. (1-4)
I staggered the notes so no channel triggered notes at the same exact time.

As far as DMG-01 with mGB, I have tweaked the 'editor'(M4LP) settings to all logical settings including extreme.
I have turned off the screen, select-a. I did no CC changes or tweaks during the test.

The best you will get is a dropped note or stuck note or errant note every 20 bars.
Obviously with everything set normally, you can experience problems in as little as every 4 bars.
It didn't matter if I only played notes in just one channel or all four simultaneously.

CGB experiences none of these problems.  I even clocked the CGB all the way down with a ltc1799 timer and
I don't have any performance issues.
Hope this helps and if anyone has any other insights to what could help, I would really appreciate it.

Should I be using sync? I wasn't writing with anything besides the DMG. Also I'm real new to MIDI in general, out side of using knobs to control VSTs. Although now it seems like the problem was using the MIDI rack instruments to control it. Or at least using that specific project file. I'll do some tests and see what works best for me. Thanks!

another option is to use the editor and change the bit and byte delay setting. Raising it should increase stability but also increase the amount of time it takes for the gameboy to get and process the data.

If the DMG gets data faster than mGB's buffer can hold this could be a symptom. Make sure you are not sending midi sync to the same midi port to cut down the bandwidth, also if you are sending CCs, dont use slopes, use steps ie: hit ctrl-b and use the pencel to draw your automation as oppose to having a sloped line "\" to limit the bandwidth. a easy test is to disable the DMG drawing routines by turning off the screen ie: select-a - if the performance is better you know you are sending too much data- or if you use a CGB or better and the performance is better as well- since mgb's serial speed will double on a CGB or better.

I just dicked around a bit, all I noticed were little bits of the arp dropping out, but generally pretty good.

Curious, do you get stuck notes or errant notes every few bars or so while sending notes to three or more channels?

Hm. Started a new project instead of using the one with the mGB midi racks someone on here made, and every problem disappeared. Weird.

It stopped working with the version I was using, so I upgraded it to see if it would fix the problem.