Ask your players how they feel about it
It's possible they don't feel any slowing down (or are ok with it). Maybe a big group isn't a problem for them.
If they want to stay as one group, ask them to help you to make the game flow better. That include:
- Knowing the rules. Not necessary all the rules, but enough so that they know what to roll most of the time.
- Knowing their character and their number. No by heart of course, but they should know where to look for that sudden Natation skill check.
- Knowing the rules specific to their character like how to use sneak attacks, what their spells do, ...
- Still being 'active' when it's not their turn, by listening to what's happening and thinking about what to do.
You can also prepare yourself before the session. For instance:
Have the battleground and the 'flow of battle' ready before the session
Have all informations quickly accessible. I personally have a paper with a short abstract of every NPC to be in the battle. All the attack and defences values, HP, equipment, special abilities... in the same place, always visible to me.
Since you're on Roll20:
- You can use the macro included in the character sheets (or build your own, or make your player build them) to make the game faster. Things like attacking or rolling for Initiative can as quick as clicking a button and reading the result.
If they also feel they're too many, you can split the party.
You can for instance play the two group in the same campaign, with one group going after one half of a McGuffin, while the other track a group of bad guys wanting to use the other half of the McGuffin to gain power beyond imagination.
An other possibility is to split the party, and run the same campaign twice. It greatly reduce your preparation load (even if they follow very different path, you can reuse assets like map and NPC without the players noticing), but you get other problems like not remembering which team did what in that village, or spoilers from one team to another. Careful note taking and a simple explanation to your players should avoid both those issues.
In both case, you need to see about your schedule. Do you run twice as many games? Will it be one group one week and the other the next one? It's for you to decide as the group.
Telegraph your decisions excessively, and allow retcons
In my experience, what the DM is thinking is very different from what the player is experiencing. In the example you gave, while you "already knew the weight limit," the player did not. From his perspective, then, you randomly collapsed the platform.
I deal with this issue by not only telling the players what the characters see, but also what the characters judge. For example, instead of telling your player "you see a crumbling platform," you can tell your player "you think the platform might collapse at any moment". In "real life," the characters have a wide range of experiences and perceptions that you can't possibly convey in a reasonable time as a DM, so telling the players what their characters think is an easy shorthand--the players are always free to disagree or probe more deeply, if they choose.
It's always going to be a judgment call on whether you should call for some knowledge-based roll or how much information to give, but I would err on the side of giving more information, to avoid instances like the example above.
Additionally, I let my characters do minor retcons. When your player decides to go on the ledge anyway, you can say something like "the ledge feels like it's about to give way under your feet". If the player changes their mind about going on the ledge, I let them do it.
It's going to be annoying, and it will feel like you're giving everything away, but it ultimately leaves your players feeling like they have a much better understanding of the situation and the logic you're using. Moreover, there might be some situations where the players will want to go on the crumbling platform--in those cases, they will have a good idea of what they're getting into.
"Gotcha" moments suck: focus more on avoiding them and less on mechanical surprises
Here's the thing: even though you're literally pitting the players against adversaries, the DM-player relationship in 5e should not be adversarial. I put in terrain and traps so that my players can have fun defeating them, not so that I can spring surprises on them and laugh evilly.
Even for actual, hidden traps, gotcha surprises are terrible. From the player's perspective, they get put into difficult, damaging situations totally out of the blue. Indeed, when I play traps, I give the players a few seconds to try to respond to a trap activation, just so they have a bit of agency ("you hear a click when you open the door, what do you do?"). By boosting player agency and letting them understand what you're thinking, you can reduce the number of gotchas and likely reduce your perceived unfairness.
Reset your player's expectations
Now, in order to implement this new DMing style, you have to fix your relationship with your players. After all, this problem is not only coming from you, but from your players as well.
You should talk to your players, tell them that you're going to change how you're DMing, and see if they're willing to reduce their combativeness. Hopefully, this "reset" will help your future sessions go more smoothly. Again, you can see this as part of making your intentions more transparent and more explicit, and hopefully your players will appreciate that.
Best Answer
I've run a game in which a turning point in the storyline was represented by a one on one boxing brawl between the party's Fighter and the local crime lord's champion (who would have won a straightforward one on one brawl).
I actually just left it to the party to come up with ways to influence this fight, which ended up being mostly underhanded. They thought of some really interesting stuff! Including a lot of distractions, subtle buffs and debuffs, stuff I could never have thought of in a hundred years of GM planning. Bribing the guards, spiking the refreshments, using illusions to summon orcish cheerleaders... Basically the party will come up with ways of using their characters to influence this showdown, all you have to do is be flexible, throw out the odd suggestion, and be ready to explain the possible consequences of failure. The deciding rolls should come down to the bard though - he is the one on the stage.
The technique you'll probably end up using will be a cumulative series of opposed checks, building to an total end goal of X successes. These checks can and should be influenced by all manner of things, and be ready to think of a few for your NPCs to use to keep it an interesting and a closely-fought thing (he busts out an amazing flute solo! advantage on his next roll!).
Use the reactions of the crowd to keep the party engaged and the tension high - do they cheer and stamp their feet, or start throwing rotten vegetables at the loser? Have a nefarious bunch bet money on the other side, and are now making death threats? Do the town guards rush over to try to break up the fight? Or one of the participant's family members?
But basically just be flexible, and let the party run riot - they will be the ones to make this a session to remember.