This also seems to be the only place mockNode.usertxns is used. Perhaps you can pull the thread a bit more and delete more stuff.

And this means AddressTxns can go away from LedgerForAPI and from data.Ledger, I think.

And I don't think:

func (tx Transaction) MatchAddress(addr basics.Address, spec SpecialAddresses) bool {

will need to take SpecialAddresses anymore.

AddressTxns is still used by the REST API (for fetching the latest state proof txn in a given round range)

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 56.05%. Comparing base (c07f2d5) to head (6aae672).
Report is 1 commits behind head on master.

@@           Coverage Diff           @@
##           master    #5983   +/-   ##
=======================================
  Coverage   56.05%   56.05%           
=======================================
  Files         481      481           
  Lines       67679    67615   -64     
=======================================
- Hits        37936    37902   -34     
+ Misses      27177    27141   -36     
- Partials     2566     2572    +6     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

AddressTxns is still used by the REST API (for fetching the latest state proof txn in a given round range)

I missed that. It's serious overkill though. We're just looking for txns from the stateproof sender. The heart of AddressTxns does this:

	var res []transactions.SignedTxnWithAD
	payset, err := blk.DecodePaysetFlat()
	if err != nil {
		return nil, err
	}
	for _, tx := range payset {
		if tx.Txn.MatchAddress(id, spec) {
			res = append(res, tx)
		}
	}

But there's really no reason to decode all of the txns. AddrTxns could become TxnsFrom(sender basics.Address) since we only need that case now, and we can check sender without decoding the stib or using MatchAddress. (MatchAddress is its own weird overkill - which a slice of addresses is built up, just to then check if id is in it)

I'm perfectly happy to approve as is, though.