Regression in broadcasting between blockarray and normal array

This used to work on 0.16.3:

a = PseudoBlockArray(zeros(10 + 12, 10 + 12), [10, 12], [10, 12]) 
b = rand(22, 22)
a .+= b .* 1.0

It currently errors with:

ERROR: DimensionMismatch("array could not be broadcast to match destination")
Stacktrace:
  [1] check_broadcast_shape
    @ ./broadcast.jl:540 [inlined]
  [2] check_broadcast_axes
    @ ./broadcast.jl:543 [inlined]
  [3] check_broadcast_axes
    @ ./broadcast.jl:547 [inlined]
  [4] instantiate
    @ ./broadcast.jl:284 [inlined]
  [5] materialize!
    @ ./broadcast.jl:871 [inlined]
  [6] materialize!
    @ ./broadcast.jl:868 [inlined]
  [7] broadcast!(::typeof(+), ::SubArray{Float64, 2, Matrix{Float64}, Tuple{UnitRange{Int64}, UnitRange{Int64}}, false}, ::SubArray{Float64, 2, Matrix{Float64}, Tuple{UnitRange{Int64}, UnitRange{Int64}}, false}, ::Base.Broadcast.Broadcasted{Base.Broadcast.DefaultArrayStyle{2}, Nothing, typeof(*), Tuple{Matrix{Float64}, Float64}})
    @ Base.Broadcast ./broadcast.jl:837
  [8] macro expansion
    @ ~/.julia/packages/BlockArrays/B2UXn/src/blockbroadcast.jl:175 [inlined]
  [9] _generic_blockbroadcast_copyto!(dest::PseudoBlockMatrix{Float64, Matrix{Float64}, Tuple{BlockedUnitRange{Vector{Int64}}, BlockedUnitRange{Vector{Int64}}}}, bc::Base.Broadcast.Broadcasted{BlockArrays.PseudoBlockStyle{2}, Tuple{BlockedUnitRange{Vector{Int64}}, BlockedUnitRange{Vector{Int64}}}, typeof(+), Tuple{PseudoBlockMatrix{Float64, Matrix{Float64}, Tuple{BlockedUnitRange{Vector{Int64}}, BlockedUnitRange{Vector{Int64}}}}, Base.Broadcast.Broadcasted{Base.Broadcast.DefaultArrayStyle{2}, Nothing, typeof(*), Tuple{Matrix{Float64}, Float64}}}})
    @ BlockArrays ~/.julia/packages/BlockArrays/B2UXn/src/blockbroadcast.jl:151
 [10] copyto!(dest::PseudoBlockMatrix{Float64, Matrix{Float64}, Tuple{BlockedUnitRange{Vector{Int64}}, BlockedUnitRange{Vector{Int64}}}}, bc::Base.Broadcast.Broadcasted{BlockArrays.PseudoBlockStyle{2}, Tuple{BlockedUnitRange{Vector{Int64}}, BlockedUnitRange{Vector{Int64}}}, typeof(+), Tuple{PseudoBlockMatrix{Float64, Matrix{Float64}, Tuple{BlockedUnitRange{Vector{Int64}}, BlockedUnitRange{Vector{Int64}}}}, Base.Broadcast.Broadcasted{Base.Broadcast.DefaultArrayStyle{2}, Nothing, typeof(*), Tuple{Matrix{Float64}, Float64}}}})
    @ BlockArrays ~/.julia/packages/BlockArrays/B2UXn/src/blockbroadcast.jl:193
 [11] materialize!(dest::PseudoBlockMatrix{Float64, Matrix{Float64}, Tuple{BlockedUnitRange{Vector{Int64}}, BlockedUnitRange{Vector{Int64}}}}, bc::Base.Broadcast.Broadcasted{BlockArrays.PseudoBlockStyle{2}, Nothing, typeof(+), Tuple{PseudoBlockMatrix{Float64, Matrix{Float64}, Tuple{BlockedUnitRange{Vector{Int64}}, BlockedUnitRange{Vector{Int64}}}}, Base.Broadcast.Broadcasted{Base.Broadcast.DefaultArrayStyle{2}, Nothing, typeof(*), Tuple{Matrix{Float64}, Float64}}}})
    @ BlockArrays ~/.julia/packages/BlockArrays/B2UXn/src/blockbroadcast.jl:148
 [12] top-level scope
    @ REPL[4]:1

Reproduce the example from the issue with PseudoBlockArray and a normal Julia array. Start in src/blockbroadcast.jl, especially the broadcast! and copyto! paths shown in the stack trace, and compare the destination and broadcast axes. Done means the a .+= b .* 1.0 operation works as it did in version 0.16.3.