Trying several remote file spellings in one ssh round trip
Instead of one ssh run per candidate path (plus a $HOME lookup for ~ and a stat), send one POSIX sh script: candidates base64-encoded in a for loop, decode each, expand a leading ~/ with $HOME on the host, and for the first [ -f ] && [ -r ] file print a header line (index, mtime size from stat -L -c "%Y %s" with a BSD stat -L -f "%m %z" fallback, base64 path) and then exec cat or head -c cap+1. Exit codes distinguish none-found and too-large. Wrap the whole script base64-encoded in sh -c eval so csh/tcsh/fish login shells run it unchanged, and search for the header at a line start so login-shell banners on stdout are skipped.
Reproduced the remote half with dash 0.5.12 and bash 5.2 (no ssh, HOME pointed at a scratch dir). Candidates ~/dir (a directory), ~/nope, ~/link (symlink to a 6-byte file): header came back as index 2, size 6, then the content, rc=0. A 100-byte file with cap=50 gave header + 51 bytes, so cap+1 lets the client detect "too large" even if the file grew after stat. All-missing gave rc=3. One gotcha when you build the header path on the host: GNU base64 (coreutils 9.4) wraps at 76 chars, so a 78-char path came out as 2 lines and splits the header. Pipe through
tr -d '\n'.