We don't hear it common but over and over there has been transfer gone wrong when sending Bitcoin to address, either to a wrong address or to a wrong network. Identifying address by format eradicate such mistakes because it helps decide what type of wallet or transaction is in use before sending.
It can not eradicate it completely. Someone can mistakenly send bitcoin to a wrong address due to clipboard malware. Address poisoning is common to altcoins but dust attack may likely lead to such also if it is bitcoin. We have seen recently someone affected with it.
The old time has gone, you can not see anyone that will say he sent bitcoin to litecoin or bitcoin cash (BCH) address again. If you want to send to litecoin address that start with 1, you will see an error that the address is not valid. In the past, people sending to the litecoin addresses that start from 3 were affected, but later litecoin developers changed it from 3 to M prefix addresses. Also you can now see that bitcoin cash addresses are starting from different characters already and not 1.
Other altcoins have different characters also.
So, if you see someone send bitcoin to a wrong address, it will definitely be because of another thing entirely like clipboard malware and address poisoning.
P2SH | Begins with "3" Still part of segwit but older version
It is pay to script hash and not part of segwit in any way, but it was used to create nested segwit (P2SH-P2WPKH) and also used in multisig wallet (P2SH-P2WSH) because segwit was not compatible at the time and some nodes has not upgraded. It has lesser fee than legacy address but I do not see any reason people should still be using it when all nodes have upgraded to segwit which has lowest fee.
If you have many inputs in your transaction and you want to consolidate or send more UTXOs (which is the output) to an address, P2TR will have the cheapest fee. If your transaction has more outputs but low inputs, segwit version 0 which you refer to as native segwit has the lowest fee. If you just want to send one input and one output, the fee slightly lower using segwit and taproot is. You can read this about fee comparison between segwit and taproot:
Pay-to-taproot (P2TR) transaction feeIn normal usage, transaction outputs are more than the inputs, which is the reason it was concluded that segwit has the lowest fee. I mean bech32.
It is worth knowing that P2TR is also segwit. bc1q is segwit version 0, while P2TR is the segwit version 1.
P2TR means pay-to-taproot.
So you can control the fee rate but size depends on your UTXOs.
It is virtual size or weight. Although, if you are referring to legacy address, it is size but the size is the same as the virtual size so far it is legacy address.
Also, it is very important to remember that taproot transactions are less supported across wallets and explore. Which can raise concern or even rejection if you are using a more advanced setups like custom scripts or PSBT. So yeah, it is the latest upgrade but it cannot be all efficient to everyone because most individuals use Bitcoin differently, some prefer custom setups.
The major advantage of taproot that was proposed was multisig which will significantly reduce transaction fee even far more than if compared to legacy and segwit multisig addresses. But I am surprised BIP86 was introduced which is single signature of taproot and we have not seen its multisig.
