Is it still considered best practice to have redundant LTMs in on datacenter if you are running on VMware?
My guess you could have an outage if you are doing an upgrade but if you existing infrastructure is tied with LTMs, maybe this isn't an issue. Just looking for best design practices.
Do people still use HA LTMs in vmware? Does it really depend on environment and requirements at this point?
Well, it definitely depends on the requirements for your company.
I would definitely still consider it best practices to have an HA pair.
Just because you are running BIG-IP virtually doesn't mean that the hardware you are running on won't fail, and that is what HA is there to help with.
I would go so far as to say if you are running virtually your hardware is more likely to fail as the hardware devices from F5 are designed specifically for longevity and have proprietary components that are known to work very well together.
If you have two physical hardware devices running VMware with a BIG-IP on each, and with them configured in an HA pair with network failover you are MUCH less likely to have an outage due to a hardware failure. Also as you mentioned you could fail over to the other device for upgrades to avoid an outage.
Hope that helps! If it does please up-vote and select this answer, it'd be greatly appreciated!