If wanting the MIDI tracks co-mingling with the audio tracks is around keeping track of the MIDI tracks: Change the default preference to embed MIDI into the project to make MIDI files. That leads to the MIDI files following your time stamp preferences. That leads to being able to easily link them up to the audio takes later on. That leads to being able to just let the MIDI files collect in the MIDI recording track out of sight until you need to grab one.
Re: Multiple audio interfaces
The 2nd step of sync'ing all sample rate clocks to a master device when making an aggregate device is the critical step. Easy for interfaces with word clock I/O or a digital audio input.
When you want to aggregate a virtual audio device or some hardware that doesn't have word clock or a digital input is when you have to try sync'ing over the data connection which is pointedly not recommended. It WILL lose sync and crash at some point!
My workflow recording with a Roland TD-50X drum kit in aggregate device with the main interface: This is one of those rogue devices with no clock input possible. (And with all the rest of the capability on this thing, Roland should feel bad!)
Aggregate device of the main interface + the Roland TD-50X
Tick the box in Audio MIDI Setup to make the main interface resample to the Roland!
[This is counter intuitive. I've found that audio interfaces often have features to resolve jittery incoming sample rate clock. That leads to the interface putting up with chasing the Roland easier than the other way around. Same deal with virtual audio devices in aggregate device.]
I record it to single 32 channel files at 32 bit floating point. The Roland spits the digital samples out at 32 bit float and you can just capture them 1:1 at any level (even above zero) and never have to soundcheck anything. Just this part alone almost makes up for the clock input faux pas.
I record a MIDI track along with in a separate track.
The MIDI track is a safety!
The audio system crashes about 1 in 50 takes from clock drift. The MIDI is still intact when this happens. I can run the MIDI back to the Roland and get the performance back. And recording single 32 channel files at 32 bit float still feels impressive or something.