We are left with following problem, upon which TcT provides the
certificate YES(?,O(n^1)).

Strict Trs:
  { group3(@l) -> group3#1(@l)
  , group3#1(::(@x, @xs)) -> group3#2(@xs, @x)
  , group3#1(nil()) -> nil()
  , group3#2(::(@y, @ys), @x) -> group3#3(@ys, @x, @y)
  , group3#2(nil(), @x) -> nil()
  , group3#3(::(@z, @zs), @x, @y) ->
    ::(tuple#3(@x, @y, @z), group3(@zs))
  , group3#3(nil(), @x, @y) -> nil()
  , zip3(@l1, @l2, @l3) -> zip3#1(@l1, @l2, @l3)
  , zip3#1(::(@x, @xs), @l2, @l3) -> zip3#2(@l2, @l3, @x, @xs)
  , zip3#1(nil(), @l2, @l3) -> nil()
  , zip3#2(::(@y, @ys), @l3, @x, @xs) ->
    zip3#3(@l3, @x, @xs, @y, @ys)
  , zip3#2(nil(), @l3, @x, @xs) -> nil()
  , zip3#3(::(@z, @zs), @x, @xs, @y, @ys) ->
    ::(tuple#3(@x, @y, @z), zip3(@xs, @ys, @zs))
  , zip3#3(nil(), @x, @xs, @y, @ys) -> nil() }
Obligation:
  innermost runtime complexity
Answer:
  YES(?,O(n^1))

The problem is match-bounded by 2. The enriched problem is
compatible with the following automaton.
{ group3_0(2) -> 1
, group3_1(2) -> 4
, group3#1_0(2) -> 1
, group3#1_1(2) -> 1
, group3#1_2(2) -> 4
, ::_0(2, 2) -> 2
, ::_1(3, 4) -> 1
, ::_1(3, 4) -> 4
, group3#2_0(2, 2) -> 1
, group3#2_1(2, 2) -> 1
, group3#2_1(2, 2) -> 4
, nil_0() -> 2
, nil_1() -> 1
, nil_1() -> 4
, group3#3_0(2, 2, 2) -> 1
, group3#3_1(2, 2, 2) -> 1
, group3#3_1(2, 2, 2) -> 4
, tuple#3_0(2, 2, 2) -> 2
, tuple#3_1(2, 2, 2) -> 3
, zip3_0(2, 2, 2) -> 1
, zip3_1(2, 2, 2) -> 4
, zip3#1_0(2, 2, 2) -> 1
, zip3#1_1(2, 2, 2) -> 1
, zip3#1_2(2, 2, 2) -> 4
, zip3#2_0(2, 2, 2, 2) -> 1
, zip3#2_1(2, 2, 2, 2) -> 1
, zip3#2_1(2, 2, 2, 2) -> 4
, zip3#3_0(2, 2, 2, 2, 2) -> 1
, zip3#3_1(2, 2, 2, 2, 2) -> 1
, zip3#3_1(2, 2, 2, 2, 2) -> 4 }

Hurray, we answered YES(?,O(n^1))