Output rows based on matching date time intervals

Do you really need second level accuracy (as in deals 321 and 654) or is minute level accuracy sufficient?

Asking because the standard answer ("Use GENERATESERIES and INTERSECT") would be much tougher to do on seconds level.

Here is an alternative approach.  In your "Modified Table A"  add a calculated column:

deal_id = 
var t = 'Modified Table A'[timestamp]
var p = 'Modified Table A'[product_id]
var d = CALCULATETABLE('Modified Table B','Modified Table B'[start_date_time]<=t,'Modified Table B'[end_date_time]>=t,'Modified Table B'[product_id]=p)
return CONCATENATEX(d,'Modified Table B'[deal_id],"")

That will give you the deal ID for any order that has a deal.

However, when trying to link the tables via the deal_id field you will (not surprisingly) encounter a circular reference error.  That means it would be better to implement that calculated column in Power Query instead.

let
    Source = Table.FromRows(Json.Document(Binary.Decompress(Binary.FromText("bc/RDYAgDEXRVUy/SewrFNFViPuvYVuCmuhfQw630DtBMiXCqquwYGkH+GD2I8lF69Z2OlMnG4OVwVANDbe3rWrJgnDuXzmLQf967t21uVZcfnI2BoMMVz0XzmImAQlnt4bT2/Gfm2uf335fd14=", BinaryEncoding.Base64), Compression.Deflate)), let _t = ((type nullable text) meta [Serialized.Text = true]) in type table [order_id = _t, timestamp = _t, product_id = _t]),
    #"Changed Type" = Table.TransformColumnTypes(Source,{{"timestamp", type datetime}}),
    #"Merged Queries" = Table.NestedJoin(#"Changed Type", {"product_id"}, #"Modified Table B", {"product_id"}, "Modified Table B", JoinKind.LeftOuter),
    #"Expanded Modified Table B" = Table.ExpandTableColumn(#"Merged Queries", "Modified Table B", {"deal_id", "start_date_time", "end_date_time"}, {"deal_id", "start_date_time", "end_date_time"}),
    #"Added Custom" = Table.AddColumn(#"Expanded Modified Table B", "Match", each if [start_date_time]<=[timestamp] and [timestamp]<= [end_date_time] then 1 else 0),
    #"Replaced Errors" = Table.ReplaceErrorValues(#"Added Custom", {{"Match", 0}}),
    #"Added Custom1" = Table.AddColumn(#"Replaced Errors", "d", each if [Match]=1 then [deal_id] else null),
    #"Removed Other Columns" = Table.SelectColumns(#"Added Custom1",{"order_id", "timestamp", "product_id", "d"}),
    #"Renamed Columns" = Table.RenameColumns(#"Removed Other Columns",{{"d", "deal_id"}}),
    #"Changed Type1" = Table.TransformColumnTypes(#"Renamed Columns",{{"deal_id", type text}})
in
    #"Changed Type1"

which then allows us to link the tables and produce the expected outcome