Solid multi-rom support with compatibility will decide this battle.
I am always wary of such claims as I have seen many and few really played out that way. Granted we now live in a post R4 world where all the kids want to be master hackers that are not inclined to lift more than a finger to help that out but hey.
Things I have seen
Download play with decide the winner. It didn't, possibly as flash carts got cheap and easy so everybody had one but suffice it to say it did not matter.
Compatibility will decide the winner. Broadly perhaps but the R4 rose up to be champion and its compatibility was based on it having updates every other day, carts like the EZ5 with a savelist that ran most games for some years after the last savelist update rapidly dropped the marginally harder to use (drag and drop R4 vs select the appropriate save or download a premade database that anybody could update) savelist until such a time as Nintendo's AP caught up (several years later) and everybody brought it back as "clean mode", "special mode", "ghost mode" or whatever they wanted to call it.
Cheats will decide the winner. Again kind of, most cheat engines were not great things compared to the likes of Nitro Hax, the emulators and things the old GBAtemp cheat crew and affiliates got sorted. That said the lag between cheats going beyond Dipstar and hackers with skills to nitrohax to action replay to on flash cart did have a little bit of an effect.
Slow motion will decide the winner. To this day things are lacking it, granted any half skilled hacker could add it to any game and the likes of DSATM will attempt to add it to any game at the press of a button.
DLDI support. Nobody was kicking up a stink about this and in fact it went the opposite way. Flash cart companies, after being prodded by homebrew types, then dragged the ROM using set kicking and screaming into it, perhaps ironically such a move is probably the biggest +1 to your "drag and drop" concept. Amusingly though before the pre DLDI systems fell apart homebrew support was a thing some took note of, much of that was probably moonshell and some nebulous "other homebrew" remark though.
SDHC support will decide the winner. Everybody more or less brought out a new cart to do it.
DSi support (in DS mode) will decide the winner. This was interesting but I have never understood the "jump ship" mentality (you have a perfectly working DS lite that will work until the DS lite or the flash cart dies, why change?) so I will leave that.
DSi support (in DSi mode) will decide the winner. Granted the DSi being a bit of a flop probably changed things but where is your cycloDS now?
Soft reset will decide the winner. Having never owned a DS lite and had to face its weak power button I have never been one for it, most added it anyway as it is easy enough to do. When soft reset support was lagging for certain carts though nobody complained too hard.
Extra features will decide the winner. This is not the DStwo extra homebrew but things like savestates and walkthroughs. Some liked it but Wood/AKAIO lacked real support for either for years.
In the end "if it can play pokemon, kingdom hearts, whatever mario platformer is the current hotness and mario kart (maybe golden sun and zelda if one of those is happening) less than 24 hours after the scene release the rest of the compatibility can be less than stellar as long as it is near braindead simple to operate" is what makes the winner. This is what the R4 (back when it was updated by the M3/R4 people) seemed to do despite it being an inferior cart in many ways to the alternatives.
All I can really say is it is not clearly superior tech that is still affordable and workable by mere mortals that decides the winner.